Variation in rehabilitation setting after uncomplicated total knee or hip arthroplasty: a call for evidence-based guidelines

This study was a planned sub-study of a large multi-centre, prospective observational cohort of people with knee or hip OA who underwent primary TKA or THA in 19 public and private Australian hospitals between August 2013 and January 2015. The study was prospectively registered in https://​clinicaltrials.​gov/​ct2/​show/​NCT01899443 (July 2013).

Results of the larger study and other sub-studies have been reported previously [12, 13, 56‐58] Initially, patients were enrolled using purposive sampling from a random selection of high-volume hospitals (performing over 275 knee or hip arthroplasties in 2012) in Australia. Later, enrolment included convenience sampling necessitated by the protracted time taken to gain departmental or executive approval using random sampling alone. The study was approved by multiple ethics committees (see Declarations for details), and written, informed consent was obtained from all participants.

Trained site coordinators screened and recruited study participants during their routine pre-operative assessment, typically within 6 weeks of surgery. The larger cohort included adults (age ≥ 18 years) with a primary diagnosis of OA in the index joint, with willingness and capacity to be followed-up by telephone post-surgery, and capacity to comprehend the study protocol, presented in English. People were excluded from the larger cohort if they were undergoing revision, hemi-arthroplasty or resurfacing surgery or had surgery planned on another joint within 3 months of the first. For the current study, we further restricted entry to people who would be eligible to undergo rehabilitation in any of the settings under consideration (home-based, facility-based or both), and therefore we excluded people who experienced a significant complication during the acute-care phase (Additional file 1), people undergoing bilateral surgeries, and self-funded participants as their insurance status changed across time; that is, for surgery they were typically coded as ‘privately insured’, but for their rehabilitation, they changed to being a ‘publicly insured’ patient. A subset of people from the larger cohort who were concurrently enrolled in a randomized rehabilitation trial at one of the participating sites were also excluded.

At the pre-admission assessment, consenting participants provided socio-demographic and anthropometric details on a study proforma. Co-morbid conditions were recorded together with any prescription medication taken daily for any stated condition. Study participants also completed patient-reported outcome measures (the Oxford Knee or Hip Score (OKS or OHS) [12, 13] and the EuroQol ‘today’ health score (EQ. 0–100 scale) [12, 13]. The OKS and OHS each present 12 Likert-style questions pertaining to index joint pain and function. Each question is scored 0–4 based on degree of symptoms or impairment; the maximal score possible is 48, with higher scores indicating better status. Similarly, for the EQ scale, higher scores indicate better perceived health status.

Site coordinators recorded acute-care outcomes, including discharge destination, reasons for referral to inpatient rehabilitation (‘patient choice’, ‘surgeon choice’, ‘lack of social support’, ‘poor progress’, ‘post-operative complication’, ‘other’), and complications on a study proforma at the time of patient discharge.

Post-discharge outcomes were collected via telephone follow-up at 35 and 90 days by trained research officers (RO). To assist recall, a study diary was provided to participants at the time of enrolment. Outcomes of interest included length of stay (LOS) in an inpatient rehabilitation unit (where applicable), number of visits to physiotherapy clinics or day hospitals or use of domiciliary services, whether supervised physiotherapy was ongoing at 90 days or whether only an unmonitored home program (defined below) was followed since discharge from acute-care. The ROs confirmed details of rehabilitation ascertained from the site, the medical record audit and/or the previous telephone call as part of the follow-up process. We considered participants lost to follow-up if their 90-day assessment was not completed within 2 weeks of the due date or after a minimum five attempts to contact them.

For quality control, study investigators re-abstracted medical record data provided by the site coordinator. Study personnel double-entered baseline, complication and discharge destination data, and corrected any disparities using data from the original patient study file.

The primary outcome, rehabilitation setting, was divided into two broad categories (facility-based or home-based) that we further split into five categories as follows:

Facility-based:

Home-based rehabilitation:

We also monitored participants’ use of community support packages, i.e., privately or publicly funded access to a limited number of services e.g. transport, meal provision, house cleaning, after discharge from the acute hospital as differences in access to such packages may explain differential reliance on inpatient rehabilitation between public and private sectors.

As this was a subset of participants from a larger study, the sample size was dictated by the sample size of the larger study and then by the number of participants who fulfilled the secondary eligibility criteria.

Following assessment of distributional assumptions, we used descriptive statistics as appropriate to describe the cohort. We compared characteristics of the study cohort and rehabilitation setting by insurance status (public or private) using unpaired, two-sided t-tests or Wilcoxon rank-sum tests for continuous variables and χ² tests or Fisher’s exact tests, as appropriate. The knee and hip cohorts were analysed separately.

We completed all analyses using Stata Version 15.1 (www.​stata.​com; StataCorp., College Station, TX), and used p < 0.05 as the cut-off for statistical significance. We followed the STROBE guidelines for reporting.

For the larger study, 1905 had surgery within the study timeframe from one of 19 hospitals (n = 9 private) (TKA n = 1074 [56%]; THA n = 831 [44%]); 1334 people (n = 621 TKA) were eligible for the sub-study after applying the additional exclusion criteria for the current study. We obtained complete data on rehabilitation to 90 days post-surgery on 1302 people (n = 610 TKA, n = 692 THA) (97%) (Fig. 1). Of the final cohort, 814 (62.5%) were privately insured, and the majority of these (n = 797, 97.9%), were operated upon in the private sector. All publicly insured participants were operated upon in the public sector.

Baseline characteristics of the cohort are shown in Table 1. Regardless of joint (hip or knee), compared to privately-insured participants, publicly-insured participants had a significantly higher BMI and worse pre-operative joint impairment (Oxford scores). Additionally, the proportion with a specific medical condition was often significantly greater amongst the publicly insured participants. Publicly-insured participants had a shorter acute hospital LOS.

Following TKA, 158 (25.9%) participated predominantly in home-based rehabilitation with greater participation observed amongst those publicly insured (32.9 vs 21.4%, P < 0.001) (Table 2). Referral to facility-based rehabilitation (inpatient rehabilitation and/or outpatient therapy) varied widely amongst hospitals. [Additional file 2: Figures S1 and S2, illustrate the variation in utilisation of the different rehabilitation settings at hospital level for both TKA and THA surgeries.] Referral to inpatient rehabilitation was more common amongst privately insured participants (56.0 (207/370) vs 7.5% (18/240), P < 0.001). Reasons for referral to inpatient rehabilitation varied by insurance status, with the predominant reason for referral for publicly insured participants being poor progress (8/18) whilst surgeon or patient choice (n = 177, [85.5%]) was the most common reason for privately insured participants (Additional file 2: Table S1). The difference in median LOS in inpatient rehabilitation between public and private participants did not reach statistical significance [Public 14 (10,21) vs Private 12 (8,14) days, P = 0.06].

Privately-insured participants who received outpatient therapy (outpatient only or inpatient followed by outpatient) had a greater number of sessions (private: median [IQR] 8 (6,12) sessions; public: 6 [5, 8]; P < 0.001) (Table 3). Privately insured participants who participated in a home program also received more domiciliary visits (median [IQR] 9 (4,10) vs 4 (3,6), P = 0.019). Additional file 2: Table S2, summarises the types of outpatient-based rehabilitation received; variation in ‘type’ of intervention was evident within and between sectors. Telerehabilitation was not reported by any participant.

At 90 days post-surgery, only 5% (12/240) of publicly-insured and 5.7% (21/370) of privately-insured patients were still receiving facility-based therapy (P = 0.72), and utilisation of community support packages was similar regardless of insurance status [Public 12.1% (29/240) vs Private 10% (37/370), P = 0.42].

Following THA, 423 participants (61.1%) participated predominantly in a home-based program with greater participation observed amongst those publicly insured (71.0 vs 55.6%, P < 0.001) (Table 2). As for TKA, inpatient rehabilitation was more common amongst privately insured participants (33.1 vs 4.4%, P < 0.001). The predominant reason for referral to inpatient rehabilitation for publicly insured participants was poor progress (5%), whilst surgeon or patient choice (71.4%) was the most common reason for privately insured participants (Additional file 2: Table S2). Median LOS in inpatient rehabilitation was longer for publicly insured participants (14 [7, 28] vs 10 [7, 14] days; P = 0.022).

Privately-insured participants who received outpatient therapy (outpatient only or inpatient followed by outpatient) had a greater number of sessions (private: median [IQR] 8 [5, 11] sessions; public: 6 [4, 8]; p < 0.001) (Table 3). As seen with TKA participants, variation in the type of outpatient therapy following THA was evident within the sector and as well as between sectors (Additional file 2: Table S2).

A minority of THA recipients were still receiving facility-based therapy at 90 days post-surgery; the proportion was higher for privately insured recipients (Public 1.2% (3/248) vs Private 4.5% (20/444), P = 0.020). Utilisation of community support packages was similar regardless of insurance status [Public 12.1% (30/248) vs Private 11.7% (52/444), P = 0.88].

To the best of our knowledge, this is the first large-scale, prospective review of what rehabilitation settings are commonly used by TKA and THA recipients. We report rehabilitation setting used across a large, clearly defined homogeneous cohort, followed prospectively with deliberate exclusion of those who were not equally eligible for home-based and facility-based rehabilitation. The characteristics of the cohort reflect those reported in previous Australian studies [16, 30, 34, 35, 38, 41, 42, 44, 46]. The higher proportion of private recipients in this cohort (62.5%) reflects that of the total arthroplasty cohort in Australia, as do our age and BMI profiles [7], thus, the patterns observed should be broadly generalisable to the greater Australian arthroplasty population. That our rates of inpatient rehabilitation utilisation amongst private hospitals align with those reported by Australia’s largest insurer support the generalisability of our data. Our observations are relevant for healthcare sectors elsewhere - such as the United States [47, 48, 50, 63, 65] and Germany [66] - where sizeable private sectors provide arthroplasty services to those privately insured. Our observations and discussion are also relevant to health systems that already rely predominantly on outpatient- and home-based programs such as the United Kingdom [25], Canada [21], and the Netherlands [24, 66] because they also highlight the variation that persists in community-based programs and between what is provided for knee and hip arthroplasty recipients, and bring to the fore the uncertainty about the appropriate duration of community-based rehabilitation following either surgery.

A limitation of the study is the reliance on patient recall to describe rehabilitation received. Whilst we incorporated several mechanisms to reduce recall bias including questioning of rehabilitation received at both 35 and 90 days post-surgery, use of patient recall diaries, and hospital-level evidence of inpatient rehabilitation admission, it is possible that utilisation of outpatient-based therapy was mis-reported. Further, we provide a snapshot of rehabilitation received (setting and number of treatments) by individuals without regard for individual needs. Whilst we acknowledge there may be some patients who require more supervised therapy – and we excluded those for whom this may be most obvious - we believe that it is unlikely that variation in individual need explains much of the variation observed. This is most apparent when we consider the lower rates of participation in facility-based therapy (both in terms of proportion of recipients and number of treatment sessions) amongst the publicly insured recipients despite their greater pre-operative joint impairment, BMI and comorbid burden, and shorter acute hospital length of stay. Our utilisation rates may not be generalisable to remote locations; no participant hospital was located in a remote region and this may explain why no participant reported the use of telerehabilitation. Finally, as the focus of this paper is on the utilisation of different rehabilitation settings, we have purposely not reported outcomes including satisfaction with the program received. We have previously reported outcomes for this cohort between those who did and did not have inpatient rehabilitation and found no differences in patient-reported joint-specific and health-related quality of life scores at 35, 90 and 365 days post-surgery [12, 13]. We acknowledge, however, that participant satisfaction with particular rehabilitation settings and programs may be contributing to the evidence-practice gap we have observed.