Feasibility of a Wearable-Based Physical Activity Goal-Setting Intervention Among Individuals With Anterior Cruciate Ligament Reconstruction

This was a single-blinded feasibility study in which all participants completed a 28-day PA observation period immediately followed by a 28-day individualized PA goal-setting intervention period. The database manager (M.P.), who was responsible for aggregating all PA-related data, was blinded to participant identity, as each person's PA monitor device was registered using a study identifier. The study was approved by the Michigan State University Institutional Review Board, and all participants provided written informed consent before the study began.

Twelve individuals with a history of primary unilateral ACLR were recruited from the Michigan State University community (Table 1). Participants were included if they were between 18 and 30 years old, had been cleared for unrestricted PA after surgery, and had no history of lower extremity surgery since their ACLR. Volunteers were excluded if they reported a history of any health condition that would limit their ability to participate in PA or increase the risk of adverse events related to participation in recreational PA. Recruits were also excluded if they had not been regularly participating in recreational PA before the injury (Tegner activity level ≥5). Individuals were not excluded if they had previous experience with PA monitoring or static PA goal setting (eg, 10 000 steps a day); however, all participants agreed to discontinue the use of other PA monitoring or goal-setting approaches during the study. Surgical characteristics and pertinent health information were captured via a patient-completed health history form at the study enrollment visit. Although this information was collected, the ACLR graft source and time since surgery were not considered inclusion or exclusion criteria in this feasibility study.

Patient-reported outcome measures were obtained via an online survey platform (Qualtrics). Participants completed the Tegner Activity Scale¹⁴  at the enrollment visit so that we could assess their level of activity to confirm adherence to the inclusion criteria. The Tegner Activity Scale ranges from 0 (sick leave or disability pension because of knee problems) to 10 (competitive sports [national elite]: soccer, football, rugby). Participants also completed the Knee Osteoarthritis Outcome Score (KOOS) at the enrollment visit and the end of the study. The KOOS was used to quantify 5 constructs: (1) knee symptom frequency and severity, (2) knee pain frequency and severity, (3) difficulty with activities of daily living, (4) difficulty with sports and recreational activities, and (5) knee-related quality of life. The Quality of Life subscale (KOOS-QOL) was the primary patient-reported outcome of interest in this investigation. The KOOS subscales have established test-retest reliability (intraclass correlation coefficients = 0.61–0.95) and minimal detectable change (MDC; 5.0–12.0 points) among individuals with knee injuries.¹⁵ 

All participants were outfitted with the Charge 3 PA monitor (Fitbit, Inc), which we selected because it is well suited for long-term monitoring based on the ease of charging, ability to wirelessly synchronize data using a smartphone, and ability to capture data using the Fitbit Application Programming Interface (API). Currently, the Charge 3 monitor has not been independently validated because of its recent commercial release (fourth quarter, 2018). Step counts assessed using the previous version of the monitor (Charge 2), which used the same accelerometer technology, were strongly correlated (intraclass correlation coefficient = 0.89) with a waist-worn monitor (model GT9X-BT; ActiGraph, LLC), which is a research-grade monitor widely used in PA measurement research.¹⁶  All monitors were registered using unique study identifiers to allow the study team and the participant to access the participant's PA data via the Fitbit mobile application (version 3.7) on their personal mobile devices. A study team member then oriented participants to the monitor features and Fitbit application and instructed each person to wear the monitor at all times unless engaged in water-based activity or recharging the monitor. Daily step-count data were captured using the Fitbit API and were stored using a customized SQL Server database (Microsoft Corp) on a secure server at Michigan State University.

Participants completed a 28-day PA observation period (commonly referred to as a run-in period) after enrollment to establish monitoring compliance and provide us with an understanding of PA patterns among our sample.¹⁷  During the observation period, participants had access to a fully functional version of the Fitbit monitor and mobile application. They were contacted on a weekly basis via a secure messaging platform (remind.com; Remind) to prompt them to wear their monitor daily and charge the monitor at least once per week. Participants were instructed to contact the study team via the secure messaging platform in the case of a technical concern or deviation from the study protocol. At the end of the observation period, participant compliance (percentage of days captured) was evaluated to determine whether the individual would remain in the study for the personalized goal-setting period. Acceptable compliance was established a priori as the successful capture of data from 80% of valid monitor wear days (>2000 steps).¹⁸  We opted for a conservative step-count cutoff for a valid day of wear as compared with other studies, in which step-count cutoffs have ranged from 300 to 1500 steps per day.¹⁸  It is important to note that monitor wear time is commonly used to assess compliance with research-grade PA monitors.¹⁹  However, wear-time data were not readily available to us via the Fitbit API, which was the rationale, in addition to support from previous authors^18,20  who have used similar commercially available monitors, for the use of minimum step counts as our primary indicator of compliance. We selected 80% of valid monitor wear days as our indicator of acceptable compliance because this value fell between previously used minimum compliance values and actual compliance values for similar feasibility studies and randomized controlled trials.^17,20–22  In the case of the values used to assess valid days of monitor wear and minimum days of monitor wear to establish monitor compliance, we opted for conservative values to maximize the volume of PA data captured despite our limited sample size and relatively short intervention period. Individuals who did not achieve acceptable compliance were allowed to keep their monitor but were not included in the personalized goal-setting period. However, it is important to note that all participants completed the KOOS at the study end point.

After the observation period, participants were transitioned to the personalized goal-setting intervention. Personalized daily step-count goals were generated based on a percentile scheduled adaptive goal-setting paradigm similar to that described by Moon et al¹³  and Adams et al.²⁰  This approach to goal setting is well established in the basic science²³  and behavioral change^20,22  literature surrounding improvement in health behaviors among sedentary adults. Daily, the preceding 9 valid days of step data were rank ordered and the step count that ranked in the 70th percentile was adopted as the step-count goal for the next day. We selected a 70th-percentile goal based on subtle modification of a previous investigation²⁰  that used a 60th-percentile goal in sedentary and obese adults. As our population of interest was in better health and therefore physically able to participate in a greater volume of PA, we opted for a slightly more aggressive goal to better match the abilities of our participants. If invalid data were recorded, the next previous valid day was included in the percentile rank process. Daily step-count goals were generated automatically using a customized script we developed. The script was scheduled to run in automated fashion at the start (12:10 am EST) of each day of data collection, and daily summaries of participants' goals were communicated digitally with us via a secure Web-based interface. The participants' individual step-count goals were transmitted via personalized message each morning (6:00–8:00 am local time) along with a supportive message based on their achievement of the personalized goal on the previous day (eg, “Great job on achieving your goal yesterday” or “Your goal is in reach today”). Participants were contacted via text message at the beginning of the week to remind them to wear their monitor daily and charge the monitor at least once per week.

The feasibility of PA monitoring and the personalized step-goal intervention were assessed based on the percentage of days of valid (>2000 steps) step-count data captured and the proportion of participants who achieved acceptable monitor compliance (>80% of days captured) during the observation and goal-setting intervention periods. We evaluated the preliminary effects of PA monitoring and the personalized step-goal intervention on the daily step count by comparing the average daily step counts captured during the observation period with those captured during the goal-setting period on an individual basis. In addition, we compared average daily step counts during each week of the observation and goal-setting intervention periods between participants who were categorized as having a high (>10 000 steps per day) or low (<10 000 steps per day) step count during the first week of the observation period.

Prestudy-to-poststudy changes in KOOS subscale scores were compared with MDC values established among athletes with a recent history of knee injury by Salavati et al.²⁴  The MDC for the KOOS-QOL subscale, our primary patient-reported outcome of interest, was 7.2 points. In the same study, MDCs for the remaining subscales were also established: Pain = 6.1 points, Symptoms = 8.5 points, Activities of Daily Living = 8.0 points, and Sports and Recreation = 5.8 points. It should be noted that these MDC values were established in a sample of competitive athletes with knee injuries, which may limit their generalizability to our sample of recreationally active adults with ACLR.²⁴  We also assessed the relationship between the prestudy-to-poststudy change in KOOS-QOL and average daily step count using Spearman ρ correlation coefficients. This was a small-scale feasibility study for which we did not generate an a priori sample-size estimate to address the primary aim. However, we did recruit a minimum number of participants to estimate the relationship between the changes in PA and KOOS-QOL score based on recommendations by Bonett and Wright.²⁵  All statistical analyses were completed using an open-source statistical package (version 0.11.1.0; JASP Team).

Among the 12 enrolled participants, average PA monitor compliance was 95.5% ± 7.3%, with individual-participant compliance ranging from 100.0% to 75.0% (Table 2). Based on the data collected during the observation period, 11 participants (91.7%) maintained acceptable compliance (>80% of days captured) and were retained in the study throughout the goal-setting period. A single participant (8.3%) was categorized as noncompliant during the observation period (75.0% compliance) and removed from the study. Among the 11 participants who were retained throughout the goal-setting period, average PA monitor compliance was 97.7% ± 2.9%, with individual compliance ranging from 100.0% to 92.9%, indicating that all retained participants achieved minimal acceptable compliance during this period. Regarding goal attainment during the goal-setting period, participants achieved their individualized step-count goals on an average of 31.5% ± 6.8% of days, with individuals ranging from 42.9% to 21.4% goal attainment over the 4-week intervention (Table 3). It should be noted that despite the relatively modest variability within the sample during the 4-week intervention, significant between-weeks variability was present within individuals' goal attainment (Figure 1).

During the observation period, participants completed an average of 9247 ± 1789 steps per day, with individual average step counts ranging from 7724 to 12 846 steps per day over this 28-day period (Table 3). During the goal-setting period, participants completed an average of 8812 ± 1068 steps per day, with individual average step counts ranging from 7678 to 10 763 steps per day. Individual weekly step-count data for the observation and goal-setting periods can be found in Figure 2. Overall, 5 of 11 participants (45.5%) displayed an increase in average daily step count from the observation period to the intervention period, with improvements ranging from 263 to 1109 steps per day. The data were divided to display activity patterns among individuals who were classified as high activity (step count >10 000 steps per day) or low activity (step count <10 000 steps per day) during the first week of the observation period.

Individual and summary statistics for the KOOS subscales assessed before and at the end of the study period are shown in Table 4. Individuals who were classified as compliant during the observation period displayed changes in the KOOS-QOL subscale that ranged from a worsening of −37.5 points to an improvement of 62.5 points. Five participants described meaningful improvements (ie, exceeded the subscale MDC) on the subscale, whereas 1 participant observed a meaningful decline (Table 4). Notably, meaningful improvement was shown by 5 participants on the Symptoms subscale, 3 participants on the Pain subscale, 2 participants on the Sports and Recreation subscale, and 1 participant on the Activities of Daily Living subscale. Finally, the change in KOOS-QOL score was not significantly related to the change in average daily step count (ρ = 0.17, P = .63).

Interventions that promote physical activity are common in the treatment of individuals with or at risk for chronic disease.^6,10  Although individuals with ACLR are less physically active^1,2  and experience a > 40% increased risk of knee joint osteoarthritis compared with peers without a history of knee surgery,⁴  interventions to promote PA have not become the standard of care in the treatment of this at-risk population. The primary purpose of our study was to evaluate the feasibility of implementing a technology-driven PA promotion intervention to deliver individualized daily PA goals to a group of individuals with a history of ACLR. Our findings indicated that participants with ACLR were highly compliant with 2 months of PA monitoring using a commercially available device (Table 2). However, when provided with individualized PA goals during the second month of the study period, they struggled to meet daily goals (Table 3), which resulted in inconsistent changes in PA participation (Figure 1). Based on these findings, it appears that long-term PA monitoring among young individuals with ACLR is feasible but that additional refinement of our approach to individualized goal setting may be required to more consistently improve their PA participation.

Compliance with PA monitoring using commercially available, wrist-worn PA monitors is well established in a variety of groups currently managing or at risk for chronic diseases such as cardiovascular disease, diabetes, and osteoarthritis.^21,26–28  To our knowledge, this is the first assessment of compliance with prolonged PA monitoring and achievement of individualized PA goals among people with ACLR. In our study, individuals with ACLR wore their PA monitor an average of 97.5% ± 1.8% of the total available days over the 56-day study period (Table 2), which is consistent with previous reports^18,29,30  of other pathologic and healthy populations who used a similar PA monitor (>90.0% compliance). During the first 28-day period, which was limited to passive PA monitoring without personalized goal setting, 11 of 12 participants (91.7%) maintained minimal acceptable monitor compliance (>80% of days captured). The single participant who did not meet minimum compliance was a 20-year-old woman who was a college student and had recently started a new internship. She indicated that the new demands on her schedule made it a challenge to remember to wear and charge her PA monitor, especially because the PA goals and charging reminder text messages often arrived after she had already left home for school in the morning. Based on these findings, it appears that continuous PA monitoring using a commercially available, wrist-worn PA monitor among young individuals with a history of ACLR is feasible, but individualizing the timing and frequency of PA goal and monitor charging reminders may be a realistic approach to improving monitor wear compliance in a larger-scale trial.

Static (eg, “take at least 10 000 steps per day”) and adaptive PA goal-setting interventions have been implemented among a variety of populations with various levels of success.^11,13  Based on this body of literature, monitoring compliance and frequency of goal achievement tend to be related to positive change in PA engagement over the course of the interventions. During the 28-day goal-setting period, the 11 compliant participants successfully wore the PA monitor on 97.7% ± 2.9% of possible days (Table 2). However, these individuals only met or exceeded their individualized, adaptive PA goal on 21.4% to 42.9% of possible days, despite excellent compliance with PA monitoring (Table 3 and Figure 1). Earlier investigators^18,20,22,31  who used adaptive goal setting among older, sedentary adults reported goal achievement rates ranging from 34.0% to 58.2%. However, these authors used significantly longer intervention periods (2–6 months) and formalized patient education throughout the intervention period, and the studies with the highest rates of achievement included a small financial incentive for meeting daily PA goals.^18,20,22,31  Additionally, we selected an aggressive 70th-percentile adaptive goal based on the nature of our relatively young patient population, which is in contrast to the 60th-percentile goals or machine learning–driven goals that have been used successfully among older, sedentary populations.²⁰  The selected percentile goal determines the level of challenge associated with each daily goal, as higher percentile goals will require more steps to achieve the daily goal. Although we thought that younger participants would be more likely to be able to achieve more aggressive goals, this was not the case, given that our goal-attainment rates were consistent with or worse than those of several studies^13,18,20  in which 60th-percentile goals were used in less-active populations. It should be noted that when participants were grouped based on baseline step counts (Figure 2), individuals who were involved in less PA at the initiation of the study did appear to experience a better response to our intervention relative to those who were classified as active at baseline. Future authors should consider including a financial incentive or gamification of goal achievement (eg, gaining points or ranking for achieving goals) to improve participant engagement and motivation. Furthermore, use of a more modest adaptive goal-setting paradigm (eg, 60th percentile) or a goal-setting paradigm that dynamically adjusts goal difficulty over the course of the intervention based on patient behavior may facilitate more consistent early success in goal achievement, which has been shown to improve long-term goal achievement and resulted in greater improvement in PA.^20,22 

The secondary purpose of our study was to evaluate the effectiveness of the individualized goal-setting intervention in enhancing patient-reported QOL in individuals with ACLR. Among participants who completed both the 28-day observation period with minimal acceptable compliance and the individualized goal-setting intervention period, 45.5% displayed improvements in average daily step count when comparing the data captured from the 28-day observation period with the 28-day intervention period; improvements ranged from 263 to 1109 steps per day. Despite this finding, we observed a median decline in average weekly step count of 2275 (range = 179–3899) steps from week 1 to week 8 of the study period. Upon reflection, we may be able to attribute this finding to participants unintentionally enrolling in a study in which the final 2 weeks of the goal-setting period corresponded with the Thanksgiving holiday or finals week for 10 participants (Table 2). Consequently, 10 participants experienced declines in daily step counts near the end of the study period based on disruptions in their normal lifestyle due to the holiday or final examinations. As a result of this unforeseen scheduling conflict, we retrospectively assessed the change in average step count between weeks 1 and 6. Interestingly, 5 of 6 participants (45.5%) with the smallest week 1 step counts experienced improvements in week 6 ranging from 113 to 3526 steps per day (Figure 2).

Regarding changes in patient-reported outcomes during the study, we found a median improvement in the KOOS-QOL score of 6.2 (range = −37.5 to 62.5) points; however, this change was not significantly related to the change in average daily step counts over the course of the study (ρ = 0.17, P = .63). Although the relationship between step counts and QOL has not been investigated among individuals with ACLR, Bell et al¹  showed in a cross-sectional study that step counts were not associated with knee-related function (Pearson r = −0.008), and Kuenze et al³²  reported that minutes of weekly PA were not associated with KOOS-QOL scores (ρ = −0.23) among individuals with ACLR who were demographically similar to those in our study. Among those who completed the goal-setting period, 5 of 11 participants described improvements in QOL that exceeded the MDC for the subscale, as opposed to a single participant who experienced a measurable decline. It is important to note that, based on the subscale scores at baseline, ceiling effects may have been a factor in several of the KOOS subscales, which limited our ability to detect changes in patient-reported outcomes during the study. For example, 3 participants demonstrated preintervention scores of 93.8 or higher on the KOOS-QOL subscale, meaning that it would have been impossible for them to report a change in the quality of life that exceeded the MDC. In addition, as indicated in the Methods section, despite being the best available estimates of MDC for the KOOS subscales, the MDC values we used were established in a sample of competitive athletes with knee injuries, which may have resulted in overestimation or underestimation of the minimal change required to exceed measurement error in a group of recreationally active individuals with ACLR.²⁴  Based on these findings, researchers designing future clinical trials to evaluate the effectiveness or efficacy of a similar PA promotion intervention should consider (1) incorporating longer goal-setting periods (2–6 months) that are consistent with those used in previous PA promotion studies in an effort to evaluate habitual behavior and minimize the effect of a single week of diminished PA; (2) using a more modest percentile-based goal calculation (eg, 60th percentile) or modifiable threshold that dynamically adjusts to participants' behaviors to ensure that the goals remain attainable; (3) including a mechanism for feedback from participants, especially during times of reduced PA, to better understand the rationale for this behavior; (4) scheduling more regular assessments of patient-reported outcome measures to determine the association between changes in PA and changes in QOL; and (5) reducing the length of the preintervention monitoring period to allow for confirmation of monitor compliance while maximizing the likelihood of participant engagement at the start of the intervention period.

Several limitations must be considered when interpreting the results of this preliminary feasibility trial. As this was a small-scale feasibility study, we did not include a control group to compare the effects of continued PA monitoring in the absence of goal setting. Future investigators may consider including a continuous PA monitoring group to establish the effects of our measurement approach on PA participation among individuals with ACLR. Our sample of individuals with ACLR was modest (N = 12) and heterogeneous based on reported times since surgery, graft sources, and current patient-reported PA levels and relatively imbalanced based on participant sex (10 women and 2 men). However, the sample characteristics were consistent with a previous finding¹  of diminished PA participation among individuals with ACLR. Regardless, it would be helpful to consider implementation of a similar intervention at a consistent time after surgery, perhaps during the terminal phases of rehabilitation, and among a sample with a better balance of sexes to establish habitual PA behaviors in a representative sample during a time when individuals may be more apt to integrate new approaches to health promotion. Lastly, given the modest goal-achievement rate among our sample (21.4% to 42.9%), it would be advantageous to include a way of determining if participants received and interacted with their personalized goal messages (eg, message-read receipts), as failure to do so would be an important factor in understanding the rationale for consistent failure to achieve step-count goals.

Step-count monitoring with concurrent individualized goal setting appears to be a feasible approach to PA promotion among young individuals with a history of ACLR. Yet based on the relatively low rates of daily goal attainment and variable change in the average daily step count over the intervention period, continued refinement of this intervention would be beneficial before implementation on a larger scale. The benefits of individualized goal setting may be more consistent if targeted toward individuals not already classified as adequately physically active, and the integration of more modest daily goals or tangible incentives for goal attainment may result in more consistent improvements in PA among this population.