Monitoring Patient Recovery After THA or TKA Using Mobile Technology

Abstract

Background

Smartphones offer the possibility of assessing recovery of mobility after total hip or knee arthroplasty (THA or TKA) passively and reliably, as well as facilitating the collection of patient-reported outcome measures (PROMs) with greater frequency.

Questions/Purposes

We investigated the feasibility of using mobile technology to collect daily step data and biweekly PROMs to track recovery after total joint arthroplasty.

Methods

Pre- and post-operative daily steps were recorded in prospectively enrolled patients (128 THA and 139 TKA) via an app, which uses the phone’s accelerometer. During 6-month follow-up, patients also completed PROMs (the pain numeric rating scale, the Hip Disability and Osteoarthritis Outcome Score Joint Replacement [HOOS JR] and the Knee Injury and Osteoarthritis Outcome Score Joint Replacement [KOOS JR]), and HOOS or KOOS JR quality of life domain via a mobile-enabled web link.

Results

At least 6 months of follow-up was completed by 65% for THA and 68% for TKA patients. Reasons for non-completion included time commitment, phone battery, app issues, and health complications. Responses from 78% of requested PROMs were returned with 96% of patients returning at least one post-operative PROM. Step data were available from 92% of days from male patients and 86% of days from female patients. The most robust recovery occurred early, within the first 2 months. The groups with higher pre-operative steps were more likely to recover their maximum daily steps at an earlier time point. Correlations between step counts and PROMs scores were modest.

Conclusion

Assessing large amounts of post-TKA and post-THA step data using mobile technology is feasible. Completion rates were good, making the technology very useful for collecting frequent PROMs. Being unable to ensure that patients always carried their phones limited our analysis of the step counts.

Electronic supplementary material

The online version of this article (10.1007/s11420-019-09746-3) contains supplementary material, which is available to authorized users.

Introduction

Mobile health (mHealth) technology, using smartphones or other mobile electronic devices, offers clinicians novel ways to interact with patients in collecting patient-centered information. Recognizing this opportunity, in 2015, the National Institutes of Health announced the Precision Medicine Initiative^® Cohort Program (now called the All of Us Research Program), which uses smartphones and other mHealth devices and applications (apps), as part of their 2020 Strategic Plan [25]. Several other groups have begun to examine the potential usefulness of mHealth for post-operative follow-up of orthopedic and other surgeries [15, 16, 18, 34, 36].

Smartphone motion sensors can capture steps taken and other mobility metrics passively and present a reliable means of assessing patients’ recovery following total hip or knee arthroplasty (THA or TKA) or other orthopedic procedures [4]. Potential advantages over traditional pedometers include better compliance with use and greater ease and capacity for data collection over longer periods in larger cohorts [1, 30]. Smartphones also offer the possibility of collecting other relevant information such as distances traveled, flights of stairs taken, or duration of activity [8].

Patient-reported outcome measures (PROMs) can also be collected efficiently using smartphones. An app or mobile-enabled website allows surveys to be administered at any time with any frequency. Delivery of the Western Ontario and McMaster (WOMAC) via mobile devices has recently been validated [2]. In 2016, we validated short-form versions of the hip and knee disability and outcomes scores, the Hip Disability and Osteoarthritis Outcome Score Joint Replacement (HOOS JR) and the Knee Injury and Osteoarthritis Outcome Score Joint Replacement (KOOS JR), which are ideal for smartphone administration [21, 22]. Changes in the timing or frequency of survey administration may improve their sensitivity or accuracy.

In this study, we used patients’ smartphones to collect daily step information following THA or TKA passively and compared their patterns of mobility recovery with changes in their responses to several validated PROMs. Our goal was to test the feasibility of using mobile technology passively to collect daily step data as a relevant domain in recovery after joint replacement. Specifically, we queried (1) what proportion of patients provided full 6-month follow-up step and PROMs information, (2) whether THA and TKA patients recovered differently, (3) if any new patterns of recovery could be discovered, and (4) how well step counts and PROMs scores were correlated.

Patients and Methods

Patients were prospectively enrolled between February 2015 and May 2016 at the time of scheduling for THA or TKA with 12 participating surgeons at the study institution. Consented, English-speaking adults, younger than 80 years of age undergoing primary unilateral THA or TKA for osteoarthritis (OA), who owned an Android or iOS (Apple iPhone) powered smartphone were eligible. Patients were excluded if they had severe health conditions incompatible with smartphone-based step count monitoring (e.g., used a wheelchair).

We enrolled 128 THA and 139 TKA (Fig. 1) patients between the ages of 27 and 80 years at least 2 weeks prior to surgery and monitored their recovery for at least 6 months post-operatively. Eligible, enrolled patients completed a minimum of 2 days of step counts pre-operatively (median [minimum, maximum]: 24 [2, 202] days) and at least one set of PROMs pre-operatively (median [minimum, maximum]: 1 PROM [1, 10]). On average, TKA patients were older and included a greater proportion of females than THA patients (Table 1).

Fig. 1.

Study recruitment and follow-up for total hip and total knee arthroplasty (THA and TKA) patients.

Table 1.

Demographic information and pre-operative steps and patient reported outcome measures (PROMs) of patients undergoing total hip (THA) or total knee (TKA) arthroplasty

Parameters	THA (N = 130)	TKA (N = 139)	p value
Age in years
Mean ± SD	59.0 ± 10.2	63.6 ± 8	0.001
Sex
Female N (%)	66 (50.80)	87 (62.60)	0.050
Pre-operative scores
Mean steps
Median (min, max)	2078 (51, 9395)	1861 (92, 7891)	0.104
Max steps
Median (min, max)	5649 (143, 22,303)	5469 (162, 16,136)	0.169
Mean pain
Mean ± SD	5.7 ± 2	5.8 ± 2.2	0.591
Mean HOOS/KOOS QOL
Mean ± SD	30.1 ± 15.7	25.3 ± 13.3	0.007
Mean HOOS/KOOS, JR
Mean ± SD	53.8 ± 13.6	50.1 ± 10.8	0.015

HOOS/KOOS QOL Hip Disability and Osteoarthritis Outcome Score/Knee Injury and Osteoarthritis Outcome Score quality of life subscale; HOOS/KOOS JR Hip Disability and Osteoarthritis Outcome Score/Knee Injury and Osteoarthritis Outcome Score Joint Replacement

The Moves™ app (no longer available as of June 2018), which used a phone’s accelerometer to count daily steps, was installed on patients’ smartphones to gather data pre-operatively and for at least 6 months of their post-operative recovery [23]. During this period, patients also completed PROMs via a mobile-enabled web link including the pain numeric rating scale (NRS), the HOOS or KOOS quality of life (QOL) domain, and the HOOS or KOOS JR. PROMs were administered weekly, 2 weeks prior and up to 12 weeks after surgery and then once every 2 weeks up to 6 months after surgery.

Statistical Analysis

Continuous variables were compared between patient groups using two-sample t tests or Wilcoxon rank-sum tests, depending upon the distribution of the data. Categorical variables were compared between patient groups using χ² or Fisher’s exact tests, as appropriate. Cox proportional hazards models were used to compare time to return to maximum pre-operative steps post-operatively between THA and TKA patients, adjusting for age, sex, pre-operative HOOS/KOOS JR score, and maximum pre-operative steps. K-means clustering, an unsupervised machine learning algorithm, was used to identify subgroups of patients with similar post-operative step count trajectories. Five criteria were used to assess the optimal number of clusters: Caliński and Harabasz, Kryszczuk variant of Caliński and Harabasz, Genolini variant of Caliński and Harabasz, Ray and Turi, and Davies and Bouldin. Outliers were removed from the THA (n = 6 outliers) and TKA (n = 0 outliers) cohorts until at least three of five criteria agreed on the optimal number of clusters. Correlation between post-operative daily step count and PROMs was assessed via bivariate linear mixed modeling. Steps and PROMs from similar timepoints were compared. Correlation coefficients are presented as point estimates with 95% bootstrap confidence intervals calculated from 1000 resamples. Statistical analyses were performed using SAS, version 9.3 (SAS Institute, Cary, NC, USA) and the R software, version 3.3.1 (R Foundation for Statistical Computing, Vienna, Austria).

Results

Of eligible patients, approximately one-quarter (n = 103, 22% THA; n = 150, 28% TKA) were ineligible because they either did not have a smartphone or their phone was not compatible with our app (Fig. 1). Among eligible patients, about one-third (n = 98, 33% THA; n = 117, 37% TKA) declined to participate. The most common reasons were that they did not usually carry their phone or were concern about too much commitment. A few patients (n = 9, 3% THA; n = 12, 4% TKA) mentioned privacy as the reason for declining to participate. Of the patients who agreed to participate but failed to provide sufficient data to be included in the analysis, more than half (n = 47, 56% THA; n = 43, 60% TKA) did not provide a reason for their lack of interest.

Of the 128 THA and 139 TKA patients who were included in the analysis, almost all (96%) completed at least one post-operative survey. Each patient completed a median of 16 (interquartile range, IQR: 10, 15) surveys. Most (78%) of the requested post-operative surveys were returned.

Patients recorded step counts, a median of 166 days (IQR:128, 177) out of 183 days. Step counts were available from a median of 92% (IQR: 71%, 98%) of post-operative days in male patients and 86% (IQR: 66%, 95%) in female patients (male versus female, p = 0.028).

About half of THA (n = 84, 50%) and TKA (n = 72, 43%) patients agreed to participate but did not complete 6-month follow-up data; most of these patients did not provide a reason for withdrawal. Technological challenges, mentioned by about 20% of patients, were the most commonly cited reason for withdrawal (Fig. 1). These included problems with the app and draining of batteries. Health complications were the least common reason for enrolled patients to withdraw.

Of the patients who completed at least one pre- and post-operative survey and provided at least two pre- and post-operative days of step data, about two-thirds of THA (n = 83, 65%) and TKA (n = 94, 68%) patients provided complete 6-month follow-up data (Fig. 1).

Recovery was faster in THA patients than in TKA patients. Half of the patients were likely to recover maximum steps by 7.6 weeks (95% CI: 5.7, 9.1) after THA compared with 10.3 weeks (95% CI: 9.0, 12.9) after TKA (Fig. 2). Recovery of maximum steps was 1.79 (95% CI: 1.34, 2.39) times more likely in THA than TKA patients (p < 0.001, Table 2 and Fig. 2). Most of the improvement in pain NRS, HOOS QOL, and HOOS JR scores and recovery of steps occurred within the first 30 days after THA (Fig. 3). Most of the improvement in pain NRS and KOOS JR scores occurred within the first 50 days after TKA, with step counts recovering in 70 days, and KOOS QOL scores continuing to improve for 150 days (Fig. 4).

Fig. 2.

Patients were more likely to achieve pre-operative steps earlier after total hip arthroplasty (THA) than after total knee arthroplasty (TKA). Kaplan–Meier survival analysis was used to estimate the probability of patients achieving their maximum pre-operative steps over 25 weeks following surgery.

Table 2.

Probability of achieving maximum pre-operative steps after total hip (THA) or total knee (TKA) arthroplasty

Probability of achieving maximum pre-operative steps	Hazard ratio (95% CI)	p value
THA vs. TKA	1.81 (1.36, 2.42)	< 0.001
Age (per 10 years)	1.20 (1.01, 1.43)	0.043
Female vs. male	0.89 (0.68, 1.18)	0.433
Pre-operative HOOS/KOOS, JR	1.00 (0.99, 1.01)	0.765

Fig. 3.

Two clusters of total hip arthroplasty (THA) patients were identified based on their post-operative recovery trajectories. a Daily step counts. b Pain NRS score. c HOOS QOL subdomain. d HOOS JR. Cluster 1 (solid line) had lower mean and maximum pre-operative steps, lower pre-operative PROM scores, and a greater proportion of females compared with cluster 2 (broken line).

Fig. 4.

Two clusters of total knee arthroplasty (TKA) patients were identified based on their post-operative recovery trajectories. a Daily step counts. b Pain NRS score. c KOOS QOL subdomain. d KOOS JR. Cluster 1 (solid line) had lower mean and maximum pre-operative steps, lower pre-operative PROM scores, and a greater proportion of females compared with cluster 2 (broken line).

Pre-operative steps were similar in THA and TKA patients (Table 1). Higher maximum pre-operative steps was predictive of earlier recovery of post-operative steps, regardless of age, sex, and pre-operative HOOS/KOOS JR score (Table 2).

Two clusters of patients were identified based on daily step count recovery. Patients in cluster 1 (about three-fourths of THA and TKA patients) had a lower pre-operative mean and maximum step count and recovered their maximum step count more slowly after surgery than patients in cluster 2 (Tables 3 and 4, Figs. 3a and 4a). Cluster 1 patients also had lower pre-operative PROM scores and included a greater proportion of females (Tables 3 and 4). After THA, a lower proportion of cluster 1 patients (86%) recovered their maximal pre-operative steps than cluster 2 (95%). Cluster 1 THA patients had slightly slower improvements in HOOS JR scores post-operatively (Fig. 3d). Cluster 1 TKA patients had slightly slower improvement in pain NRS post-operatively (Fig. 4b).

Table 3.

Pre-operative steps and PROMs for THA patients cluster

Parameters	Cluster		p value
	1 (N = 55)	2 (N = 20)
Age in years
Mean ± SD	62.0 ± 9	63.2 ± 9.5	0.627
Sex
Female N (%)	29 (52.70)	8 (40.00)	0.330
Pre-operative scores
Mean steps
Median (Q1, Q3)	1717 (959, 2587)	4301 (3346, 5885)	< 0. 001
Max steps
Median (Q1, Q3)	4164 (2494, 7943)	10,168 (6444, 12,836)	< 0.001
Mean pain
Mean ± SD	5 ± 2	5 ± 2	0.5906
Mean HOOS QOL
Mean ± SD	33 ± 17	36 ± 12	0.484
Mean HOOS, JR
Mean ± SD	53 ± 15	509 ± 12	0.157

*Six outliers were excluded from analysis

Table 4.

Pre-operative steps and PROMs for TKA patients cluster

Parameters	Cluster		p value
	1 (N = 62)	2 (N = 29)
Age in years
Mean ± SD	63.5 ± 8.1	63.1 ± 8.2	0.835
Sex
Female N (%)	46 (74.2)	14 (48.3)	0.015
Pre-operative scores
Mean steps
Median (Q1, Q3)	1543 (682, 2213)	3789 (2896, 5152)	< 0. 001
Max steps
Median (Q1, Q3)	4498 (2157, 6836)	8480 (7246, 11,551)	< 0.001
Mean pain
Mean ± SD	6 ± 2	6 ± 2	0.262
Mean KOOS QOL
Mean ± SD	25 ± 14	24 ± 14	0.633
Mean KOOS, JR
Mean ± SD	48 ± 11	53 ± 10	0.047

Correlation between step number and PROM scores were modest (Table 5).

Table 5.

Correlation between daily step count and PROMs following THA or TKA

	THA	TKA
Versus daily steps*	Correlation coefficient (95% CI)
Pain	− 0.12 (− 0.22, 0.01)	− 0.23 (− 0.29, − 0.17)
HOOS/KOOS QOL	0.10 (− 0.02, 0.19)	0.21 (0.15, 0.27)
HOOS/KOOS, JR	0.17 (0.07, 0.26)	0.25 (0.19, 0.31)

*Correlations with daily step counts from 0 to 6 months post-operatively

Discussion

We tested the feasibility of monitoring the recovery of THA and TKA patients by collecting step counts passively and administering PROMs through patients’ smartphones. Response rates were good, with most requested PROMs completed and step counts recorded for nearly all days for 6 months post-operatively. Mobile digital technology made it possible to collect and analyze this large amount of data, revealing some interesting patterns: a faster recovery in THA than TKA patients, a period of rapid recovery in the first 2 post-operative months, and the possible presence of two clusters of patients with different recovery trajectories. Surprisingly, correlations between step counts and PROM scores were only modest.

Our enrollment, response, and completion rates compared well with those in previous studies that used digital technologies for follow-up after orthopedic surgery [15, 36, 38, 39]. About a quarter of patients were not eligible for our study because they did not have a smartphone or their phone was incompatible with our app. In comparison, 30 to 35% of Americans did not own a smartphone during our recruitment period [29], suggesting that our urban population represents an overestimation of the proportion of patients who could be followed with a smartphone, for clinical or research purposes. Our study may also be slightly skewed toward patients who are more savvy with digital devices than the general population. However, in 2019, smartphone ownership rose to 80% [29], reflecting the rapidly intensifying familiarity with digital devices in more and more Americans. This rising proportion likely includes older patients, even though in our study, those over 80 years old were excluded to increase the probability of smartphone use. Future studies, in which wider smartphone ownership would make a greater proportion of patients eligible, are needed to confirm that our observations on post-arthroplasty recovery were not biased by focusing on the 75% of patients with a smartphone compatible with our app.

Despite having smartphones, some patients declined to participate because they did not always carry them. Even among patients who did enroll, we could not ensure that they always carried their phones post-operatively. That women recorded a lower percentage of days with steps than men (86% vs 92%, respectively) may reflect women’s tendency to carry their phone in a purse, resulting in step counts only when they left their homes. We also could not capture movement during physical therapy or other exercise (e.g., running or swimming). However, the step counts we recorded were similar to those found in previous studies using traditional pedometers [10, 17, 20, 24, 32, 35, 38] or wearable devices [11, 39], providing confidence in our approach. A limitation of our study was that we chose an arbitrary 2-week pre-operative baseline, but health apps are now present on most phones, making longer-term information available, including data recorded prior to study enrollment. Overall, researchers should keep in mind that smartphones may be most useful in monitoring patients’ return to baseline behavior (including steps and phone use), rather than their gains in post-operative mobility.

About one-third of patients did not complete the study. Almost half of those patients did not give a reason for withdrawing, and 20% indicated the commitment was too great. Querying patients about their interactions with our surveys could have helped us understand whether a lower frequency of PROMs administration or other adjustment might have improved completion rates. On the other hand, recent technological improvements may address other reasons patients gave for withdrawing: smartphones now have native step counter apps, and phone batteries have improved. A few patients (less than 5%) mentioned privacy as the reason for not participating. Indeed, data security remains an unsolved issue for all mHealth applications. A recent study reported security gaps in almost all mHealth applications, making users susceptible to receiving false or inaccurate information, leakage of personal information to third parties, or health identity theft [7].

We elucidated some notable recovery patterns. Of particular interest for THA and TKA research and for post-operative clinical care, we observed that recovery occurred earlier than expected: within 30 days after THA and 50 to 150 days after TKA. A few previous studies have reported a similar timeline [13, 14, 39]. PROMs scores changed little between 6 and 12 months, suggesting that to test any interventions intended to accelerate recovery, PROMs would need to be administered earlier and with greater frequency than the current standard, possibly using mobile technology. Our observations also argue for early post-arthroplasty rehabilitation. Current protocols vary widely in timing and duration, with no consensus about which approaches are best [9, 19, 26, 33]. THA patients return to baseline step counts and improve their PROM scores earlier than TKA patients, which is consistent with previous reports [3, 5, 13, 14, 27, 28].

We also discovered two clusters of patients characterized by their post-operative step counts. PROMs alone would not have sufficed to discern this divergent pattern. The more slowly recovering cluster 1 included more women, consistent with previous studies reporting lower step counts in female THA patients [17, 35]. As mentioned above, phone-carrying behavior may have affected our findings. Clusters may represent different rates of “return to pre-operative behavior,” rather than differences in mobility recovery per se.

The modest correlation between step counts and PROMs was unexpected. A previous study reported that patients overestimated their early recovery in self-reported outcome measures, when compared with objective measures of mobility [37]. Alternatively, step counts and other objective activity measures may assess different recovery domains than PROMs. Previous studies report different degrees of correlation between step counts or activity measures and Lower Extremity Activity Scale (LEAS) [34], UCLA activity scale [10], WOMAC scores [6], and other PROMs [12].

Our study demonstrates the feasibility of tracking post-THA and post-TKA recovery using mobile technology and highlights potential uses and limitations. Administering PROMs with frequency had a high response rate and yielded new insight into recovery, which occurred earlier than with standard PROM administration schedules. However, because ours was a feasibility study, future studies will need to corroborate our observations using clinically important covariates such as co-morbidities, other joint involvement, or discharge disposition. Although our study supports the feasibility and utility of the mHealth approach post-arthroplasty, we also identified aspects of this technology that warrant improvement. For example, we could not ensure that patients always carried their phones with them, especially during exercise. Although only a few of our patients stated privacy as an issue, data security remains a major concern. Furthermore, managing the data from several hundred patients in our study was challenging because the technology was not designed for aggregating data across patients but rather for self-monitoring individuals’ own physical activity. As Ramkumar et al. [31] observed in a recent review, integrating a refined data capture and processing stream into the patient care pathway will require continued development of secure information infrastructures, not yet available in the healthcare marketplace.

Funding Information

Stephen Lyman, PhD, was supported in part by funds from Weill Cornell Medicine, Grant No. 5UL1TR000457.

Compliance with Ethical Standards

Conflict of Interest

Chisa Hidaka, MD, Kara Fields, MS, and Wasif Islam, BS, declare that they have no conflicts of interest. Stephen Lyman, PhD, reports personal fees from Omni, Inc., JOSKAS, and Universal Research Solutions; editorial board membership at JBJS Statistics, outside the submitted work; and support for the current work in part by funds from the Clinical Translational Science Center (CTSC), National Center for Advancing Translational Sciences (NCATS) grant #5UL1TR000457. The content is solely the responsibility of the authors and does not necessarily represent the official views of the funding source NCATS, based in Rockville, MD. David Mayman, MD, discloses personal fees from Smith and Nephew, OrthAlign, and Imagen, outside the submitted work.

Human/Animal Rights

All procedures followed were in accordance with the ethical standards of the responsible committee on human experimentation (institutional and national) and with the Helsinki Declaration of 1975, as revised in 2013.

Informed Consent

Informed consent was obtained from all patients for being included in this study.

Required Author Forms

Disclosure forms provided by the authors are available with the online version of this article.