Associations Between Physical Therapy Visits and Pain and Physical Function After Knee Arthroplasty: A Cross-Lagged Panel Analysis of People Who Catastrophize About Pain Prior to Surgery

Abstract

Objective

Physical therapy visit number and timing following knee arthroplasty (KA) are variable in daily practice. The extent to which the number and timing of physical therapy visits are associated with current and future pain and function—and, alternatively, whether pain and function are associated with the number of future physical therapy visits following KA—are unknown. The purpose of this study was to determine temporal and reciprocal associations between the number of physical therapy visits and future pain and function in people with KA.

Methods

A cross-lagged panel design was applied to a secondary analysis of data from a randomized clinical trial of patients with pain catastrophizing. The 326 participants underwent KA and completed at least 7 of 9 health care diaries over the year following surgery. The Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) pain and physical function subscales were completed preoperatively and multiple times during follow-up. Separate cross-lagged panel analyses were conducted for WOMAC pain and physical function.

Results

From surgery to 2 months postsurgery, reciprocal associations were generally not found between physical therapy visit number and future pain or function. From 2 to 6 months postsurgery, a greater number of physical therapy weekly visits were associated with higher (worse) 6-month pain and function. Higher (worse) WOMAC pain at 2 and 6 months led to more visits from 2 to 6 and 6 to 12 months, respectively.

Conclusions

Higher pain scores 2 months postsurgery were associated with higher physical therapy use in the 2 to 6 months following surgery. However, patients with increased physical therapy use from 2 to 6 months had significantly higher pain scores 6 months postsurgery. Those patients with persistent pain 6 months postsurgery were higher users of physical therapy 6 to 12 months postsurgery. This reciprocal positive association between pain and physical therapy during this time period suggests minimal benefit of physical therapy despite an increased physical therapy use for patients with higher pain.

Impact

This is the first study to determine the association between the number and timing of physical therapy visits and current and future pain and function. Based on the results, physical therapy might not be a cost-effective strategy to treat patients with persistent pain following KA.

Introduction

Costs associated with physical therapy visits following knee arthroplasty (KA) have received increased attention, particularly with the advent of cost reduction models like bundled payments developed by the Centers for Medicare and Medicaid Services.¹ This model incentivizes surgeons and hospitals by providing reimbursement based on health care services delivered over a 90-day period from the day of surgery. The greater the cost savings (ie, less care visits), relative to an established standard, the better the reimbursement for services rendered.

Substantial variation exists in the number of physical therapy visits following hospitalization. Mitchell and colleagues² reported that 709 Medicare-insured patients who received physical therapy in a clinic owned by the treating surgeon averaged 16 (SD = 9.0) visits, whereas 3062 patients treated in non–surgeon-owned practices averaged 7 (SD = 8.0) visits. Average length of time for the physical therapy episode of care was 2 months following KA. Oatis and colleagues³ examined the physical therapy records of 112 patients following KA and reported a mean (SD) of 15 (8.4) physical therapy visits, though duration of the episode of care was not reported.

We recently reported the 1-year costs of postoperative care for persons with KA.⁴ Participants in our study were recruited for a randomized clinical trial of a perioperative pain coping skills intervention designed for persons with moderate to high levels of pain catastrophizing.⁵ We found that physical therapy visits extended beyond the first 2 months following surgery and, for some patients, physical therapy continued up to 1 year following surgery. The average number of physical therapy visits over the first 90 days following surgery was 23; over the 1-year postoperative period, patients averaged 33 (SD = 23.3) visits, costing an average of $3823 (SD = $2861).⁴

Studies examining the number of physical therapy visits following KA have not examined whether these visits are associated with subsequent pain or function or whether pain and function are associated with future physical therapy use. One might suspect that as patients demonstrate improved pain and function, the number of future physical therapy visits would decrease because the need for physical therapy would be reduced as pain and function improve. In contrast, if pain is staying the same or improving at a slower rate, the need for future physical therapy visits could stay the same or increase.

Our overall purpose was to examine relationships between physical therapy visits, pain, and function. The primary objective of our study was to use a cross-lagged panel analysis⁶ to determine if the number of physical therapy visits was associated with future pain and function and vice versa. By using a cross-lagged panel analysis, we could estimate the directional association between physical therapy visits and Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) pain and physical function subscale scores as they impact each other over time while accounting for patients’ previous physical therapy visits and WOMAC scores. We hypothesized that we would find reciprocal temporally driven associations that would demonstrate the following patterns: earlier pain and function severity would predict future physical therapy visit number, and earlier physical therapy visit number would predict future pain and function severity.

Methods

Participants

The current study is a secondary analysis of panel (longitudinal) data collected during the KASTPain trial, a study of recovery following KA surgery for patients who catastrophize about pain. The treatments were pain coping skills training, arthritis education, or usual care. Findings indicated no evidence of differences between the 3 treatment arms for all primary and secondary outcome measures. The number of physical therapy visits was not controlled in the study. The KASTPain trial is thoroughly summarized in both the protocol⁷ and the trial publication.⁵

Briefly, inclusion criteria required participants to have knee osteoarthritis, have moderate to high levels of pain catastrophizing, as indicated by Pain Catastrophizing Scale⁸ scores of 16 or higher, and have a scheduled KA within 8 weeks of screening. Participants were excluded if they were scheduled for revision surgery or an additional arthroplasty surgery within 6 months of the planned KA.

Trained data collectors, blinded to treatment group, collected preoperative data in-person and follow-up data by telephone (using up to 8 attempts) at 2 weeks postsurgery and at 2, 6, and 12 months following the KA. The trial was conducted at 5 sites (Duke University, New York University Medical Center, Virginia Commonwealth University, Wake Forest University, and Southern Illinois University). The institutional review boards from all sites approved the study and all participants signed a consent form. The trial was registered at ClinicalTrials.gov (NCT01620983).

Outcome Measures for the Current Study

The focus of the current study was on postoperative WOMAC pain and WOMAC physical function subscale scores, and the number and timing of combined home-based and outpatient clinic-based physical therapy visits. The highly reliable and valid WOMAC pain and physical function subscales, 3.1 Likert version, were used to assess participants’ pain with activity and difficulty with activity, respectively.⁹ WOMAC scores were obtained preoperatively, at 2 weeks following surgery and at 2, 6, and 12 months following surgery. WOMAC pain subscale scores range from 0 to 20, whereas WOMAC physical function subscale scores range from 0 to 68. Higher scores for WOMAC pain and physical function subscales indicate worse pain with activity and worse functional difficulty, respectively.

The health care diary, used to measure the number and timing of physical therapy visits, was designed by the investigators to capture self-reported weekly data on a variety of health care encounters, each over a 6-week period. Evidence supports the validity of patient-completed diary data for capturing outpatient provider visits over a 1-year period.¹⁰ To collect a full year of posthospitalization health care data, 9 diaries were completed with the final diary covering a 4-week period. Among the data collected were weekly home-based or clinic visits for physical therapy. Study coordinators regularly contacted participants to assist with completion of the diaries over the 1-year postoperative period. To align with the timing of WOMAC measurements, the average number of weekly physical therapy visits was calculated from the time of surgery to 2 weeks postsurgery, from 2 weeks postsurgery to 2 months postsurgery, from 2 months postsurgery to 6 months postsurgery, and from 6 months postsurgery to 12 months postsurgery. We required a minimum of 7 of 9 complete diaries to capture at least 80% of the 1-year time period for the diaries to reasonably reflect the full year of data collection, and therefore excluded participants who completed less than 7 diaries. Demographic variables of age, sex, body mass index (kg/m²), race (percentage African American vs non–African American), and the Pain Catastrophizing Scale⁸ were collected at baseline.

Statistical Analysis

Sample characteristics are reported using means and SDs for continuous variables, and counts and percentages for categorical variables for the entire sample (N = 384), and for those who completed at least 7 of 9 health care diaries (n = 326) to allow for comparisons between the full and analyzed sample.

A cross-lagged and autoregressive path model¹¹ was used to assess the reciprocal relationship between WOMAC scores and average weekly physical therapy visits while accounting for temporality of effects, with separate models for WOMAC pain and WOMAC physical function subscales. Figure 1 shows an illustration of the WOMAC pain model. Our main interests were the effects of WOMAC scores at a given time point on the number of future average weekly physical therapy visits, and the number of average weekly physical therapy visits at a given time point on future WOMAC scores. All analyses were adjusted for patients’ baseline age, sex (female vs male), race (African American vs other), and Pain Catastrophizing Scale score.⁸

Figure 1.

Autoregressive and cross-lagged associations for Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) pain subscale scores and physical therapy (PT) visit number over the course of the 1-year study. The coefficients and corresponding P values are included with each association tested in the model. The model was adjusted for covariates, age, sex, race, and pain catastrophizing.

The fit of each model was assessed via the χ² test, the root-mean-square error of approximation, the standardized root-mean-square residual, and the comparative fit index. Models were deemed to have good fit if the χ² test was nonsignificant (P > .05), the root-mean-square error of approximation was < 0.05, the standardized root-mean-square residual was < 0.08, and the comparative fit index was > 0.95.¹² Models that did not fit the data were respecified until adequate model fit was achieved, as is common with structural equation modeling. Missing data were imputed via the mice package,¹³ and all analyses were performed in R 3.6.1 (R Foundation). An α of .05 was used with the path models fit using lavaan within the semTools package.¹³

Analyses were completed on the full sample and a subsample of participants who had no additional surgeries during the 12-month follow-up. The full sample analyses reflect current practice in the most generalizable way, including patients with additional surgery during follow-up. The smaller sample reflects a more pure group without additional surgeries, which would have likely led to a more challenging recovery. We compared and contrasted an all-comers approach to a pure sample of only those patients without additional surgery.

Role of the Funding Source

The funders played no role in the design, conduct, or reporting of this study. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Results

A total of 384 participants underwent KA surgery, and of these, 326 (84.9%) completed at least 7 of the 9 health care diaries and were included in the current analyses. The age, sex, body mass index, race, and pain catastrophizing scores of the 326 participants were similar to the entire surgical sample, as were the WOMAC pain and physical function scores. For those who completed a sufficient number of health diaries, the mean age was 63.2 years, 67% were female, and 35% were African American. Improvements were seen in both WOMAC pain and physical function mean (SD) scores from 11.3 (3.3) and 36.6 (11.3) at baseline to 3.1 (3.6) and 11.2 (12.6) 12 months postsurgery, respectively (Tab. 1). Almost all participants (92.3%) had at least 1 average weekly physical therapy visit between the time of surgery and the 2 weeks following surgery, whereas 29.4% of participants used physical therapy from 6 to 12 months following surgery (Tab. 2). The average number of physical therapy visits per week declined across the study period from 2.6 (SD = 1.3) visits per week from the time of surgery to 2 weeks postsurgery to 0.2 (SD = 0.4) visits per week from 6 to 12 months after surgery (Tab. 2).

Table 1.

Participant Characteristics^a

Characteristic	Entire Surgical Sample (N = 384)	Study Sample (At Least 7/9 Diaries Completed) (n = 326)
Age, y	63.2 (8.0)	63.2 (8.0)
Sex, female vs maleb	257 (67)	218 (67)
Body mass index, kg/m2	32.3 (6.2)	32.3 (6.0)
Race, % African American vs other	135 (35)	114 (35)
Pain Catastrophizing Scale score	30.0 (9.3)	30.0 (9.1)
Preoperative WOMAC pain score	11.4 (3.4)	11.3 (3.3)
2-wk WOMAC pain score	9.7 (3.8)	10.0 (3.7)
2-mo WOMAC pain score	6.1 (3.8)	6.2 (3.8)
6-mo WOMAC pain score	4.1 (3.7)	4.1 (3.7)
12-mo WOMAC pain score	3.1 (3.7)	3.1 (3.6)
Preoperative WOMAC physical function score	37.1 (11.5)	36.6 (11.3)
2-wk WOMAC physical function score	32.3 (14.0)	35.2 (13.6)
2-mo WOMAC physical function score	20.0 (12.6)	20.3 (13.0)
6-mo WOMAC physical function score	14.2 (12.1)	14.6 (12.1)
12-mo WOMAC physical function score	11.0 (12.6)	11.2 (12.6)
Good WOMAC pain outcome classb	323 (84)	274 (84)
Good WOMAC physical function outcome classb	316 (82)	267 (82)
Participants with additional follow-up surgeryb		68 (28)

^a Data are reported as mean (SD) unless otherwise indicated. WOMAC = Western Ontario and McMaster Universities Osteoarthritis Index.

^b Data are reported as number (percentage) of participants.

Table 2.

Physical Therapy Visits: Descriptive Statistics

Time Frame	No. (%) of Patients With > 0 Average Weekly No. of Physical Therapy Visits	Mean (SD) Weekly No. of Physical Therapy Visits for Patients With > 0 Average Weekly No. of Physical Therapy Visits
Surgery to 2 wk	301 (92.3)	2.8 (1.1)
2 wk to 2 mo	319 (97.9)	2.1 (0.9)
2 mo to 6 mo	245 (75.2)	0.9 (0.7)
6 mo to 12 mo	96 (29.4)	0.6 (0.6)

Results from the WOMAC pain final model are displayed in Figure 1. As expected, almost all autoregressive coefficients linking current average weekly physical therapy visits with future average weekly physical therapy visits (or current with future WOMAC scores) were statistically significant, indicating that previous levels of average weekly physical therapy visits (or WOMAC scores) are associated with their future levels; these associations were accounted for in all models. Because WOMAC pain and physical function findings were similar, we focus on WOMAC pain here in the article, and provide WOMAC physical function findings in the online supplement (Suppl. Tabs. 1 and 2; Suppl. Figs. 1 and 2).

Results for the Full Sample

Statistically significant cross-lagged effects of average weekly physical therapy visits and WOMAC pain scores were found from the 2-month postsurgery time point onward (Tab. 3; Fig. 1). In general, after 2 months postsurgery, higher WOMAC scores were associated with more weekly physical therapy visits, and more weekly physical therapy visits were associated with higher WOMAC scores. For example, WOMAC pain scores reported 2 months postsurgery were positively associated with the average number of weekly physical therapy visits from 2 to 6 months postsurgery (β₃₂ = 0.035; P < .001). This indicates that for every WOMAC pain point increase, weekly visits are expected to be higher by approximately 0.04 visits per week. For example, after accounting for previous levels of pain, a patient with a WOMAC pain score of 10 at 2 months would have averaged 0.2 greater weekly visits from 2 to 6 months postsurgery (ie, approximately 3.6 additional visits over the period) as compared with a patient with a WOMAC pain score of 5 at 2 months postsurgery. For the effect of current average weekly physical therapy visits on WOMAC pain scores, for every additional weekly visit from 2 to 6 months, WOMAC pain scores at 6 months were predicted to be higher by 0.63 points, after accounting for prior pain and physical therapy visits. WOMAC physical function results appear online.

Table 3.

WOMAC Pain Cross-Lagged Model for the Full Sample^a

Outcome	Time	Predictor	Estimate (SE)	95% CI	Standardized Estimate	z	P
WOMAC pain subscale score	Baseline	Age	−0.054 (0.021)	−0.094 to −0.014	−0.131	−2.629	.009
		Sex (female vs male)	0.710 (0.339)	0.047−1.374	0.102	2.097	.036
		Race (African American vs other)	1.245 (0.347)	0.564−1.926	0.181	3.585	<.001
		Pain Catastrophizing Scale score	0.126 (0.018)	0.091−0.161	0.349	6.979	<.001
	2 wk	Baseline WOMAC pain score	0.296 (0.067)	0.164−0.428	0.267	4.402	<.001
		Surgery to 2 wk, average no. of physical therapy visits	−0.024 (0.145)	−0.309 to 0.261	−0.009	−0.164	.87
		Age	−0.037 (0.025)	−0.086 to 0.011	−0.082	−1.499	.134
		Sex (female vs male)	−0.200 (0.407)	−0.998 to 0.599	−0.026	−0.490	.624
		Race (African American vs other)	1.433 (0.431)	0.589 to 2.276	0.188	3.327	.001
		Pain Catastrophizing Scale score	−0.033 (0.023)	−0.078 to 0.013	−0.082	−1.400	.161
	2 mo	Baseline WOMAC pain score	0.246 (0.066)	0.116−0.376	0.211	3.715	<.001
		2-wk WOMAC pain score	0.276 (0.055)	0.167−0.384	0.262	4.984	<.001
		Surgery to 2 wk, average no. of physical therapy visits	−0.166 (0.167)	−0.494 to 0.162	−0.058	−0.994	.32
		2 wk to 2 mo, average no. of physical therapy visits	0.253 (0.238)	−0.213 to 0.719	0.061	1.064	.287
		Age	−0.021 (0.024)	−0.068 to 0.027	−0.044	−0.861	.389
		Sex (female vs male)	−0.582 (0.397)	−1.360 to 0.196	−0.072	−1.465	.143
		Race (African American vs other)	1.663 (0.422)	0.837 to 2.489	0.207	3.945	<.001
		Pain Catastrophizing Scale score	0.004 (0.022)	−0.040 to 0.048	0.01	0.185	.853
	6 mo	2-wk WOMAC pain score	0.080 (0.053)	−0.025 to 0.184	0.079	1.497	.134
		2-mo WOMAC pain score	0.404 (0.053)	0.299−0.508	0.422	7.58	<.001
		2 wk to 2 mo, average no. of physical therapy visits	−0.245 (0.206)	−0.649 to 0.159	−0.062	−1.187	.235
		2-6 mo, average no. of physical therapy visits	0.613 (0.274)	0.076−1.151	0.119	2.236	.025
		Age	−0.046 (0.022)	−0.089 to −0.004	−0.101	−2.154	.031
		Sex (female vs male)	−0.167 (0.359)	−0.871 to 0.537	−0.021	−0.464	.642
		Race (African American vs other)	0.735 (0.393)	−0.035 to 1.504	0.096	1.871	.061
		Pain Catastrophizing Scale score	0.029 (0.019)	−0.009 to 0.067	0.073	1.517	.129
	12 mo	2-mo WOMAC pain score	0.160 (0.047)	0.068−0.252	0.17	3.396	.001
		6-mo WOMAC pain score	0.517 (0.051)	0.416−0.617	0.527	10.084	<.001
		2–6 mo, average no. of physical therapy visits	0.300 (0.212)	−0.115 to 0.716	0.06	1.416	.157
		6–12 mo, average no. of physical therapy visits	0.454 (0.394)	−0.319 to 1.226	0.049	1.152	.25
		Age	−0.000 (0.018)	−0.036 to 0.035	−0.001	−0.020	.984
		Sex (female vs male)	0.356 (0.302)	−0.236 to 0.948	0.047	1.179	.238
		Race (African American vs other)	0.497 (0.331)	−0.152 to 1.145	0.066	1.501	.133
		Pain Catastrophizing Scale score	0.039 (0.016)	0.008−0.070	0.1	2.473	.013
Average no. of weekly physical therapy visits	Surgery to 2 wk	Baseline WOMAC pain score	−0.007 (0.025)	−0.056 to 0.043	−0.016	−0.263	.793
		Age	0.024 (0.009)	0.006−0.043	0.145	2.54	.011
		Sex (female vs male)	0.296 (0.156)	−0.009 to 0.602	0.105	1.901	.057
		Race (African American vs other)	0.040 (0.162)	−0.278 to 0.357	0.014	0.245	.806
		Pain Catastrophizing Scale score	0.008 (0.009)	−0.010 to 0.025	0.052	0.861	.389
	2 wk to 2 mo	Surgery to 2 wk, average no. of physical therapy visits	0.389 (0.033)	0.325−0.452	0.558	11.952	<.001
		2-wk WOMAC pain score	0.016 (0.012)	−0.009 to 0.040	0.062	1.268	.205
		Age	−0.000 (0.006)	−0.011 to 0.011	−0.002	−0.037	.971
		Sex (female vs male)	0.012 (0.091)	−0.166 to 0.190	0.006	0.135	.893
		Race (African American vs other)	−0.159 (0.096)	−0.346 to 0.028	−0.082	−1.662	.096
		Pain Catastrophizing Scale score	−0.001 (0.005)	−0.011 to 0.008	−0.010	−0.218	.828
	2 mo to 6 mo	Surgery to 2 wk, average no. of physical therapy visits	−0.125 (0.030)	−0.185 to −0.066	−0.235	−4.145	<.001
		2 wk to 2 mo, average no. of physical therapy visits	0.451 (0.043)	0.367 to 0.535	0.587	10.523	<.001
		2-mo WOMAC pain score	0.035 (0.010)	0.017−0.054	0.19	3.718	<.001
		Age	0.004 (0.004)	−0.004 to 0.013	0.046	0.952	.341
		Sex (female vs male)	0.076 (0.071)	−0.062 to 0.215	0.051	1.084	.279
		Race (African American vs other)	0.126 (0.076)	−0.022 to 0.274	0.085	1.67	.095
		Pain Catastrophizing Scale score	0.004 (0.004)	−0.004 to 0.011	0.048	0.994	.32
	6 mo to 12 mo	2 wk to 2 mo, average no. of physical therapy visits	−0.058 (0.024)	−0.105 to −0.010	−0.137	−2.378	.017
		2−6 mo, average no. of physical therapy visits	0.211 (0.032)	0.148−0.274	0.387	6.539	<.001
		6-mo WOMAC pain score	0.019 (0.006)	0.007−0.031	0.178	3.038	.002
		Age	0.002 (0.003)	−0.003 to 0.007	0.034	0.648	.517
		Sex (female vs male)	0.033 (0.042)	−0.049 to 0.115	0.04	0.781	.435
		Race (African American vs other)	−0.019 (0.045)	−0.107 to 0.068	−0.024	−0.435	.664
		Pain Catastrophizing Scale score	−0.003 (0.002)	−0.007 to 0.001	−0.069	−1.309	.19

^a = 17.835 (P = .164); root-mean-square error of approximation = 0.034; standardized root-mean-square residual = 0.020; comparative fit index = 0.994. WOMAC = Western Ontario and McMaster Universities Osteoarthritis Index.

Results for Patients With No Additional Surgery

We excluded 68 participants who had 1 or more surgeries during the 1-year follow-up that might have led to increased average weekly physical therapy use. There were 41 contralateral KAs or hip replacements, 7 KA revision surgeries, 20 knee manipulations under anesthesia, and 9 orthopedic surgeries of other body regions. Compared with participants with only KA surgery over the 1-year period, participants with additional musculoskeletal surgery had worse 12-month mean (SD) WOMAC pain scores (4.3 [4.2] vs 2.7 [3.4]; t₃₁₅ = 3.1; P = .002) and a greater number of average weekly physical therapy visits (46 [28.3] vs 29 [19.8]; t₃₂₄ = 5.6; P < .001) over the 12-month period. For this pure sample, 55 participants had at least 1 physical therapy visit after 6 months postsurgery.

Cross-lagged associations remained positive between average weekly physical therapy visits and WOMAC scores for this pure subsample. However, the magnitude of many of these associations was diminished compared with the primary analysis and, in some cases, were nonsignificant (Tab. 4; Fig. 2; Suppl. Tabs 1 and 2; Suppl. Figs. 1 and 2). The only statistically significant effects were WOMAC scores 2 months postsurgery predicting average weekly physical therapy visits 2 to 6 months postsurgery (WOMAC pain: β₃₂ = 0.032; P = .003) and 6-month WOMAC on-average weekly physical therapy visits 6 to 12 months postsurgery (WOMAC pain: β₄₃ = 0.013; P = .012). WOMAC physical function results appear online. After respecification, all models fit well for the full sample and subsample analyses, according to their χ², root-mean-square error of approximation, standardized root-mean-square residual, and comparative fit index.

Table 4.

WOMAC Pain Sensitivity Analysis Cross-Lagged Model for the Pure Sample^a

Outcome	Time	Predictor	Estimate (SE)	95% CI	Standardized Estimate	z	P
WOMAC pain score	Baseline	Age	−0.090 (0.023)	−0.135 to −0.046	−0.221	−3.957	<.001
		Sex (female vs male)	0.813 (0.377)	0.074–1.553	0.117	2.156	.031
		Race (African American vs other)	1.407 (0.390)	0.643–2.170	0.204	3.611	<.001
		Pain Catastrophizing Scale score	0.096 (0.020)	0.057–0.135	0.271	4.796	<.001
	2 wk	Baseline WOMAC pain score	0.288 (0.076)	0.139–0.438	0.259	3.793	<.001
		Surgery to 2 wk, average no. of physical therapy visits	−0.083 (0.164)	−0.404 to 0.238	−0.030	−0.506	.613
		Age	−0.023 (0.028)	−0.079 to 0.033	−0.051	−0.816	.414
		Sex (female vs male)	−0.288 (0.459)	−1.188 to 0.612	−0.037	−0.627	.531
		Race (African American vs other)	1.533 (0.484)	0.584–2.482	0.200	3.165	.002
		Pain Catastrophizing Scale score	−0.027 (0.026)	−0.078 to 0.023	−0.070	−1.059	.290
	2 mo	Baseline WOMAC pain score	0.332 (0.072)	0.191–0.473	0.282	4.611	<.001
		2-wk WOMAC pain score	0.292 (0.062)	0.171–0.413	0.276	4.748	<.001
		Surgery to 2 wk, average no. of physical therapy visits	−0.178 (0.172)	−0.515 to 0.160	−0.062	−1.032	.302
		2 wk to 2 mo, average no. of physical therapy visits	0.173 (0.246)	−0.308 to 0.654	0.042	0.705	.481
		Age	−0.018 (0.026)	−0.070 to 0.034	−0.037	−0.682	.496
		Sex (female vs male)	−0.331 (0.425)	−1.165 to 0.502	−0.041	−0.779	.436
		Race (African American vs other)	1.888 (0.462)	0.982–2.794	0.233	4.084	<.001
		Pain Catastrophizing Scale score	−0.017 (0.023)	−0.063 to 0.029	−0.041	−0.735	.462
	6 mo	2-wk WOMAC pain score	0.084 (0.064)	−0.040 to 0.209	0.084	1.325	.185
		2-mo WOMAC pain score	0.353 (0.064)	0.227–0.479	0.371	5.490	<.001
		2 wk to 2 mo, average no. of physical therapy visits	−0.267 (0.248)	−0.753 to 0.219	−0.067	−1.076	.282
		2-6 mo, average no. of physical therapy visits	0.260 (0.331)	−0.389 to 0.090	0.049	0.785	.432
		Age	−0.042 (0.025)	−0.091 to 0.007	−0.091	−1.673	.094
		Sex (female vs male)	−0.071 (0.411)	−0.876 to 0.734	−0.009	−0.174	.862
		Race (African American vs other)	1.137 (0.460)	0.236–2.037	0.147	2.473	.013
		Pain Catastrophizing Scale score	0.036 (0.022)	−0.008 to 0.079	0.090	1.587	.112
	12 mo	2-mo WOMAC pain score	0.149 (0.048)	0.055–0.243	0.171	3.112	.002
		6-mo WOMAC pain score	0.481 (0.055)	0.374–0.589	0.524	8.763	<.001
		2-6 mo, average no. of physical therapy visits	0.180 (0.231)	−0.273 to 0.634	0.037	0.779	.436
		6–12 mo, average no. of physical therapy visits	0.225 (0.631)	−1.012 to 1.463	0.018	0.357	.721
		Age	−0.013 (0.020)	−0.052 to 0.025	−0.032	−0.687	.492
		Sex (female vs male)	0.572 (0.321)	−0.058 to 1.201	0.080	1.781	.075
		Race (African American vs other)	0.465 (0.373)	−0.266 to 1.197	0.066	1.247	.212
		Pain Catastrophizing Scale score	0.034 (0.017)	0.000–0.067	0.093	1.978	.048
Average no. of weekly physical therapy visits	Surgery to 2 wk	Baseline WOMAC pain score	−0.015 (0.029)	−0.072 to 0.041	−0.037	−0.524	.600
		Age	0.026 (0.011)	0.005–0.048	0.157	2.391	.017
		Sex (female vs male)	0.148 (0.177)	−0.198 to 0.495	0.052	0.841	.400
		Race (African American vs other)	0.015 (0.185)	−0.348 to 0.378	0.005	0.081	.935
		Pain Catastrophizing Scale score	0.016 (0.010)	−0.003 to 0.035	0.110	1.642	.101
	2 wk to 2 mo	Surgery to 2 wk, average no. of physical therapy visits	0.349 (0.038)	0.274–0.423	0.503	9.161	<.001
		2-wk WOMAC pain score	0.008 (0.015)	−0.021 to 0.037	0.032	0.549	.583
		Age	0.002 (0.007)	−0.011 to 0.015	0.018	0.315	.753
		Sex (female vs male)	0.008 (0.106)	−0.201 to 0.217	0.004	0.074	.941
		Race (African American vs other)	−0.080 (0.114)	−0.302 to 0.143	−0.041	−0.700	.484
		Pain Catastrophizing Scale score	0.000 (0.006)	−0.011 to 0.011	0.002	0.043	.965
	2 mo to 6 mo	Surgery to 2 wk, average no. of physical therapy visits	−0.097 (0.031)	−0.158 to −0.035	−0.186	−3.092	.002
		2 wk to 2 mo, average no. of physical therapy visits	0.449 (0.044)	0.363–0.536	0.597	10.164	<.001
		2-mo WOMAC pain score	0.032 (0.011)	0.011–0.052	0.175	3.014	.003
		Age	0.004 (0.005)	−0.005 to 0.014	0.050	0.939	.348
		Sex (female vs male)	0.090 (0.076)	−0.058 to 0.238	0.061	1.187	.235
		Race (African American vs other)	0.124 (0.083)	−0.040 to 0.287	0.084	1.481	.139
		Pain Catastrophizing Scale score	0.004 (0.004)	−0.004 to 0.012	0.057	1.075	.283
	6 mo to 12 mo	2 wk to 2 mo, average no. of physical therapy visits	−0.021 (0.020)	−0.059 to 0.018	−0.071	−1.045	.296
		2–6 mo, average no. of physical therapy visits	0.135 (0.026)	0.083–0.186	0.352	5.117	<.001
		6-mo WOMAC pain score	0.013 (0.005)	0.003–0.024	0.185	2.511	.012
		Age	−0.001 (0.002)	−0.005 to 0.003	−0.039	−0.642	.521
		Sex (female vs male)	0.043 (0.033)	−0.022 to 0.107	0.075	1.298	.194
		Race (African American vs other)	−0.033 (0.036)	−0.104 to 0.038	−0.059	−0.914	.361
		Pain Catastrophizing Scale score	−0.003 (0.002)	−0.007 to 0.000	−0.116	−1.891	.059

^a = 13.483 (P = .411); root-mean-square error of approximation = 0.012; standardized root-mean-square residual = 0.019; comparative fit index = 0.999. WOMAC = Western Ontario and McMaster Universities Osteoarthritis Index.

Figure 2.

Autoregressive and cross-lagged associations in the sensitivity analysis for Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) pain scores and physical therapy (PT) visit number over the course of the 1-year study. The coefficients and corresponding P values are included with each association tested in the model. The model was adjusted for covariates, age, sex, and race.

Discussion

This is the first study, to our knowledge, that examines reciprocal temporal associations between average weekly physical therapy visits and patient self-reported outcomes over an extended period of recovery. Our hypotheses were partially supported, because almost effects were present beginning 2 months after surgery. In general, we found that after 2 months following surgery, a higher number of weekly physical therapy visits (or worse/higher WOMAC scores) predicted higher downstream WOMAC scores (or a greater number of downstream physical therapy weekly visits). This finding illustrates that patients attending more average weekly physical therapy sessions have higher future pain and lower future functioning.

Although the current study is observational by design and does not account for unmeasured confounding,¹⁴ these analyses account for the temporality of associations between average weekly physical therapy visits and future WOMAC pain and physical function scores, and vice versa, a component of causality. This approach, in our view, has potential for informing potentially causal and reciprocal temporal associations between dosage of average weekly physical therapy visits and outcomes.

Higher WOMAC Pain Predicting Greater Number of Future Physical Therapy Visits

We found that worse pain at the 2-month (or 6-month) time point predicted a higher number of average weekly physical therapy weekly patient visits during the 2- to 6-month interval (or 6- to 12-month interval). This finding is not surprising, given that pain is well known to be one of the key drivers of average weekly physical therapy use in people with KA. Patients seek out KA because of pain, and patients with more severe pain would logically receive a higher dosage of future average weekly physical therapy visits.

As evidenced by the coefficients in Tables 3 and 4, associations (eg, a standardized coefficient of 0.190 for 2-month WOMAC pain predicting 2- to 6-month average weekly physical therapy weekly visit number) suggest that additional associated future physical therapy visit costs for people with worse WOMAC scores are sometimes notable. For example, a patient with a WOMAC pain score of 10 compared with a patient with a WOMAC pain score of 5 at the 2-month time point would incur an average of approximately 3.6 additional average weekly physical therapy visits during the 2- to 6-month time interval, resulting in approximately $415 of additional direct medical costs⁴ for the patient. Six-month WOMAC pain scores were positively associated with 6- to 12-month average weekly physical therapy visits, indicating that patients with higher (worse) pain had more visits in the 6- to 12-month time period.

Physical therapists should anticipate greater average weekly physical therapy use, particularly for those patients whose pain is persistent, because this increased use could occur up to 12 months following surgery. The challenge for physical therapists is that greater use also leads to greater costs, and, given that physical therapy is already known to account for the majority of medical costs during the year following KA,^4,15 it is likely to be a target for future cost reductions. Therefore, future research should focus on methods for enhancing efficiency of care, particularly for those patients with persistent pain, because these patients are likely the highest users of physical therapy following KA and consequently the most costly.

Higher Number of Weekly Physical Therapy Visits Predicting Future Higher WOMAC Pain Scores

A less intuitive finding was the positive association between higher visit number and higher (worse) future WOMAC scores. A higher number of average weekly physical therapy visits between 2 and 6 months predicted worse WOMAC pain scores at 6 months. For example, a patient who averaged 2.5 visits a week during the 2- to 6-month interval had 6-month WOMAC pain scores that were approximately 1.2 points higher at 6 months compared with a patient averaging 0.5 visits over the same period. These effects are small and, although statistically significant, might not be clinically meaningful unless visit rates are high during the 2- to 6-month interval. These 2- to 6-month physical therapy visit findings suggest that patients who receive substantial treatment during the 2- to 6-month interval have persistent pain and functional loss and therefore might experience less benefit during this period compared with the period from baseline to 2 months. Therefore, more physical therapy visits from 2 months onward did not associate with better future outcomes. We note that only 29.4% of our sample reported average weekly visits during the 6- to 12-month interval. With this said, the fact that 29.4% of our sample were receiving physical therapy between 6 and 12 months following surgery is a novel finding. Studies describing the duration of the physical therapy episode of care following KA indicate that physical therapy duration is typically 2 to 3 months following surgery^2,16 Our data suggest a substantial number of patients, albeit those who catastrophize about their pain, continue to receive physical therapy for at least a year following surgery.

Physical Therapy Visits and WOMAC Scores During the Early Postoperative Period

Standard of care for postoperative physical therapy is typical during the first few months following KA, because this early postoperative period is generally the time frame when physical therapy is considered critical for facilitating surgical recovery and preventing long-term impairment.^2,16,17 During the first 2 months following surgery, most patients in the current study received physical therapy 2 to 3 times per week (Tab. 2), with the average number of weekly visits tapering off after that.

We suspect the lack of significance during this time period was the case because a substantial majority of our patients received physical therapy approximately 2 to 3 times per week over this period, consistent with standard of care. Additionally, most patients in our study demonstrated substantial improvements in pain and function (Tab. 1) over this early postoperative period, a consistent finding in this population.^18,19 There was also relatively low variation in average weekly physical therapy visits combined with substantial pain and function improvement over this period, which contrasts with longer-term (ie, > 2 months following surgery) findings when average weekly physical therapy visits were more variable and likely targeted to a larger proportion of patients with persistent pain and reduced function.

Excluding Patients With No Additional Surgery Changed Reciprocal Relationships Between Pain, Function, and Weekly Physical Therapy Visits

Patients undergoing KA often undergo additional procedures, related or unrelated to the involved knee, during the year following KA.²⁰ Additional surgery is likely because multiple joints are commonly involved in persons with osteoarthritis.²¹ To enhance generalizability in our first analyses, we included all patients, regardless of the occurrence of additional surgery.

The second analysis excluded participants who underwent additional musculoskeletal procedures (n = 68) and provides information about the cross-lagged associations when examining a pure sample with only KA. We found that higher WOMAC scores were associated with higher future average weekly physical therapy visits, much as we found for the full sample. This indicates that higher WOMAC scores are a consistent driver of a higher number of future average weekly physical therapy visits, even when no additional procedures beyond the KA are conducted. In contrast, the effect of current average weekly physical therapy visits on future WOMAC pain scores was nonsignificant across all comparisons over all time periods. In other words, a higher average visit number did not associate with lower (better) or higher (worse) WOMAC scores.

These comparative findings suggest that participants who had additional surgical procedures along with subsequently greater case management complexity also were likely to have higher WOMAC pain scores over time. These cases could have been more challenging for physical therapists to treat given that other body regions required treatment. In some cases, the surgical knee could have been compromised by additional surgical intervention (eg, manipulation under anesthesia or revision surgery). We speculate that these additional procedures created greater challenges in treatment delivery and consequently less benefit over time for the surgical knee.

Strengths and Limitations

Our study has several strengths, including a large sample size, comprehensive follow-up over a 1-year period, an accounting for temporality of effects and important covariates, and a 90% follow-up of the original sample. Multiple imputation allowed us to use data from all participants to generate estimates for each time point. However, this study also has limitations. First, our sample was recruited because they had moderate to high levels of pain catastrophizing, a psychological health construct characterized by difficulty coping with pain,⁸ which could limit generalizability to other patients undergoing TKA. With this said, 1-year pain and function outcomes were similar to those reported for heterogeneous samples of patients,^18,19,22 suggesting that pain catastrophizing was not an important prognostic indicator. Second, response bias could be inherent in the weekly number of physical therapy visits, because these measures were self-reported. Third, the number of participants with visits more than 6 months after surgery was relatively small (n = 96 in the full sample and n = 55 in the pure sample), but use of maximum likelihood estimation during the analysis included all participants. Additionally, the SEs and CI width for each variable were similar across time points, suggesting reasonably stable estimates for all analyses. We also note that there is an effect of WOMAC pain at 6 months on physical therapy visits from 6 to 12 months, which indicates sufficient variability in physical therapy visits from 6 to 12 months to garner a statistically significant effect. Fourth, approximately 20% of our participants (68/326) had additional musculoskeletal surgery, which likely led to additional average weekly physical therapy visits. With this said, patients with an isolated KA still averaged approximately 30 visits over the study period.

Conclusions

From surgery to 2 months postsurgery, the number of physical therapy visits is generally not associated with future pain or function severity. Consistent improvement usually occurs in the first 2 months as the patient recovers from surgery, and a higher physical therapy dosage (ie, a higher number of weekly visits) does not appear to be critical in this early stage. From 2 to 6 months, a higher dosage of physical therapy visits associated with higher future pain and worse function. Analysis of physical therapy visit data suggests that a higher number of physical therapy visits from 2 to 6 months postsurgery is associated with minimal future benefit. Patients with persistent pain 6 months postsurgery were higher users of physical therapy 6 to 12 months postsurgery. This reciprocal positive association between pain and physical therapy during these time periods suggests minimal benefit despite an increased use of physical therapy. Approximately 20% of patients undergoing KA have a poor outcome with minimal improvement in pain and function relative to preoperative levels.²³ Our data suggest these patients are likely to receive a large number of physical therapy visits and incur substantial costs, despite a poor outcome. Our study supports the need for future research to determine causes and cost-effective care approaches for patients with persistent pain following KA.

Author Contributions

Concept/idea/research design: C.M. Orndahl, R.A. Perera, D.L. Riddle

Writing: C.M. Orndahl, R.A. Perera, D.L. Riddle

Data collection: D.L. Riddle

Data analysis: C.M. Orndahl, R.A. Perera

Project management: D.L. Riddle

Fund procurement: D.L. Riddle

Providing facilities/equipment: D.L. Riddle

Providing institutional liaisons: D.L. Riddle

Consultation (including review of manuscript before submitting): C.M. Orndahl, R.A. Perera

Acknowledgments

We thank the participants of this study. We also thank the physical therapists and nurses for their roles in providing the interventions, as well as study staff and study psychologists at Duke University, New York University, Southern Illinois University, Virginia Commonwealth University, and Wake Forest University.

Ethics Approval

The institutional review boards from Duke University, New York University Medical Center, Virginia Commonwealth University, Wake Forest University, and Southern Illinois University approved the study, and all participants signed a consent form.

Funding

This study was supported by a grant from the National Institute of Arthritis and Musculoskeletal and Skin Diseases (NIAMS) of the National Institutes of Health (NIH) (UM1AR062800). Support also was provided by a National Institutes of Health Clinical and Translational Science Awards grant from the National Center for Advancing Translational Sciences to Virginia Commonwealth University (UL1TR000058). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Clinical Trial Registration

The trial was registered at ClinicalTrials.gov (NCT01620983).

Disclosures

The authors completed the ICMJE Form for Disclosure of Potential Conflicts of Interest and reported no conflicts of interest.