Research Review By Dr. Demetry Assimakopoulos©


Download MP3

Date Posted:

November 2020

Study Title:

Digital Care for Chronic Musculoskeletal Pain: A 10,000 User Longitudinal Observational Study


Bailey JF, Agarwal V, Zheng P et al.

Author's Affiliations:

Department of Orthopaedic Surgery, University of California, San Francisco, CA; Quantitative Sciences Unit, School of Medicine, Stanford University, Palo Alto, CA; Hinge Health, Inc., San Francisco, CA; Division of Physical Medicine & Rehabilitation, Stanford University, Palo Alto, CA; Department of Physical Medicine & Rehabilitation, Vanderbilt University Medical Center, Nashville, TN.

Publication Information:

Journal of Medical Internet Research 2020; 22(5): e18250.

Background Information:

The global burden of MSK pain has been well documented. Conservative, non-surgical care, including exercise, education and behavioural coaching are universally recommended as first-line treatments (1), as they can achieve similar outcomes to surgical intervention, with lower costs and less risk (2, 3). Unfortunately, non-surgical care carries significant barriers to implementation, such as low patient engagement, lack of funding resources and the need for ongoing monitoring.

Provision of conservative MSK care through digital formats may improve outcomes for these services by increasing patient engagement and enabling patients to take a proactive role in their health care by learning self-management strategies. Although digital care programs (DCP) for chronic MSK pain have been conducted in the past, these studies are marred by low sample size and limited implementation in real-world settings. As such, the authors of this study sought to analyze the effect of DCPs for chronic low back and knee pain, in a large sample of patients within a real-world setting.

Pertinent Results:

A total of 10 264 participants were included – 6468 with back pain and 3796 with knee pain. The average participant age was 43.6 years and the average BMI was 30.25 kg/m2. Fifty-percent of participants were female. The proportion of female participants was 3.9% higher in the knee pain pathway.

73% (n = 7497) of participants completed the program, defined as those who completed at least one exercise session or read one educational article in weeks 9-12. Older users were more likely to complete the DCP (OR = 1.037; CI= 1.03-1.04). Those with higher BMI were less likely to complete (OR = 0.973; CI = 0.97-0.98). No other baseline measures were associated with completion. On average, completers engaged in 10.45 weeks with 35.02 exercise therapy sessions and 19.39 education sessions. No adverse effects were reported, apart from temporary discomfort.

The estimated mean (average) reduction in pain by week 12 was 68.45% (30.89 points on the VAS). The mean change in pain scores per week (adjusted for gender, pathway, baseline age, BMI, anxiety and depression scores) were 15.96 points for week 1 (P < 0.001) and 1.11 points per week for weeks 6-12 (P < 0.001). Changes were not significant between weeks 2-5. The minimally important change from baseline pain (defined as either a VAS reduction of 20-points or a 30% change from baseline) was achieved by 78.6% of completers, and 69.6% of all participants. Completers also demonstrated greater pain reduction than non-completers, with an increased average reduction rate of 0.48 points per week throughout weeks 2-5. The final pain reduction was nearly identical for both female and male genders, but there was a significantly higher average reduction rate for male participants in the first week, and lower mean reduction levels in weeks 2-5, and 6-12. The low back pain pathway was associated with a higher mean reduction rate than the knee pain pathway in the first week, but was not significant in later weeks.

A greater level of exercise therapy engagement was associated with greater reduction in VAS pain score (P < 0.001). Interestingly, the session-by-session rate of pain reduction (adjusted for gender, pathway, baseline age, BMI, anxiety and depression) for the initial 10 exercise therapy sessions was 1.9 VAS points per session. The number of weekly coach interactions was also associated with a reduction in pain with a mean reduction of 0.18 VAS points per interaction (P = 0.003), for the first 30 interactions. While the number of participant-to-coach interactions was associated with a mean reduction VAS pain score of 0.30 points per-interaction for the first 20 interactions, the number of coach-to-participant interactions was not significantly associated with pain reduction.

The average baseline Patient health Questionnaire (PHQ-9) score was 12.01 (moderate depression) and decreased by 57.0% to a mean score of 5.05 at week 11 (P < 0.001; mild depression). There were no differences in depression scores between pathways. A baseline PHQ-9 score > 5 (mild depression) was the threshold for patients demonstrating depressive symptoms. The percentage of patients with depressive symptoms at the beginning of the study was 21.1% and reduced to 11.4% by the end of the study. Very similar findings were shown for anxiety. The average baseline Generalized Anxiety Disorder (GAD-7) score was 11.49 (moderate anxiety) and decreased by 58.3% to a mean score of 4.78 at week 11 (P < 0.001). Back pain pathway participants demonstrated a 0.46 point (P < 0.001) greater mean reduction in GAD-7 scores than those in the knee pathway. The threshold for having anxiety symptoms was a score > 5 on the GAD-7 (mild anxiety). The percentage of patients with anxiety symptoms at baseline and at the end of the study was 28.3% and 14.2%, respectively.

The average surgery likelihood score decreased by 67.4% (8.15 points; P < 0.001), and by 66.8% and 68.2% in the knee and low back pain pathways, respectively. The mean Knee Injury and Osteoarthritis Outcome Score Pain Subscale (KOOS-Pain) score decreased by 33.9% (5.19 points; P < 0.001), while the mean Modified von Korff (MvK) score decreased by 51.4% (8.20 points; P < 0.001). The mean Work Productivity and Activity Impairment scale (WPAI) score decreased by 63.94% from baseline (20.29 points; P < 0.001). The DCP final satisfaction score was 8.97/10.

Three specific responder clusters were identified. Participants with a high pain score at baseline, with a rapid decrease during the course of the study, were labelled as “high rapid” (HR) responders. Similarly, patients with high pain at baseline, with gradual reductions in pain scores, were labelled as “high gradual” (HG). Conversely, patients with low baseline pain scores below 50/100, with a gradual response, were labelled “low gradual” (LG). HR participants had the highest mean pain reduction during the duration of the DCP (48.8 points), followed by the HG group (33.3 points), and the LG group (15.5 points). Females had higher odds of a HG response (total 17.3% of all females). The odds of a HG response were also increased by 3.1% (P < 0.001) per unit increase of BMI and by 2.2% (P = 0.001) and 2.1 (P = 0.002) per unit increase in PHQ-9 and GAD-7.

Clinical Application & Conclusions:

This retrospective cohort study was the first of its kind to demonstrate a positive effect of a 12-week digital care program (DCP) on chronic low back and knee pain in a large sample of real-world patients. Overall, patients experienced a 68.5% average improvement in VAS score between baseline and 12-weeks, and 78.6% of program completers (equates to 69.6% of all participants) achieved clinically meaningful improvement (pretty impressive!). Approximately 73% of participants completed the program – a very high completion rate! There was a significant relationship between engagement and pain reduction; specifically, completers engaged in a total of 35.0 exercise therapy sessions, 19.4 education sessions and 91.5 coaching interactions during the DCP. The authors believe that the exceptional completion rate of the study may be secondary to the multi-pronged strategy of the DCP.

The authors observed an average 68.5% improvement in VAS pain scores during the DCP, which outperforms pain reductions demonstrated in similar trials. Not only did the study demonstrate improvements in pain scores, but also showed improvements in functional outcome measures for knee pain and low back pain, and in the measured psychosocial variables of depression and anxiety.

The authors also discovered symptom response clusters: High Pain Rapid Responders (HR), High Pain Gradual Responders (HG) and Low Pain Gradual Responders (LG). The authors were able to forecast with 76% accuracy from baseline data which of these pain response groups a user would fall into based on their baseline information. The HR was more likely to occur in male participants, with lower BMI, or lower depression or anxiety scores (Writer’s note: interestingly, no description about the other responder types was elucidated).

Overall, it appears that provision of rehabilitation and coaching for low back pain and knee pain can be delivered virtually through a Digital Care Program. Continued research into response patterns may ultimately allow for a more personalized approach to care, including more accurate prognosis and additional treatment options for patients likely to have more gradual recovery.

Study Methods:

These authors implemented a retrospective cohort study design. They recruited participants with low back or knee pain through employers across the United States. Recruited employees were diverse and included both office and service-based roles. Each person participated in one of two digital care pathways (DCPs): one for chronic knee pain and the other for chronic low back pain. The only differences between the pathways were the specific exercise regimes and some condition-specific education materials (i.e. anatomy, surgical options etc.).

Inclusion criteria:
  • Ages 18-80
  • At least 12-weeks of low back of knee pain
  • Baseline VAS > 0/10
  • Completion of at least 1 exercise session or reading 1 educational article in the first two weeks following registration
Exclusion criteria:
  • Presence of red flag symptoms, including fracture, joint instability, infection, cancer and cauda equina syndrome.
All participants received a tablet computer with necessary applications installed, alongside two Bluetooth, wearable motion-sensors with straps. Instructions on how to place the motion sensory straps above and below the painful region during the in-app exercise therapy were also included. In the lower back program, a sensor was placed on the posterior low back and anterior chest, while in the knee program, sensors were placed over the anterior tibia and thigh. Sensors utilized standard accelerometer and gyrometer technology and were used to monitor compliance and technical performance of exercises.

Exercise therapy sessions included light intensity stretching and strengthening exercises commonly used in clinical practice, administered using animations and instructional videos. The app displayed real-time graphics showing the position of the participant’s relevant body parts based on the wearable sensors.

Users were also assigned a personal coach with whom they communicated on a weekly basis via text message, email or in-app messaging. Each participant was also offered a 20-30 minute phone call on 3 occasions during the DCP. Each week, participants were instructed to complete at least 3 sessions of sensor-guided exercise therapy, read two educational articles and log their symptoms at least twice. Behavioural therapy topics were addressed through educational articles and brief interactive modules, and focused on common CBT topics such as catastrophizing, active coping, fear avoidance and goal setting/tracking.

The primary outcome was the score on a pain visual analogue scale (VAS) over a 24-hour period immediately after an exercise session. Participants could optionally report VAS up to two more times weekly. The authors defined a minimally important change as a 30% or 20/100 point decrease from baseline.

Secondary Outcomes were collected at 6-weeks and 12-weeks, and included:
  • Depression: Patient Health Questionnaire-9 (PHQ-9)
  • Anxiety: Generalized Anxiety Disorder-7 (GAD-7)
  • Disability for knee pain pathway: Work Productivity and Activity Impairment (WPAI) scale and the Knee Injury and Osteoarthritis Outcome Score Pain Subscale (KOOS-Pain)
  • Disability for the low back pain pathway: Modified von Korff scale (MvK)
  • Surgery likelihood: patients were asked “What do you think the chances are you will have surgery in the next year in %.”
Baseline measures included BMI, age and gender. Participation in the DCP was measured by recording the number of exercise therapy sessions completed, the number of coaching interactions and number of education articles read. Post-program satisfaction was measured at week-12, by asking the question: “On a scale of 0-10, how likely is it that you would recommend the program to a friend of colleague?”

Study Strengths / Weaknesses:

  • Delivery of the DCP was through a company called Hinge Health. Two of the study authors were academic collaborators for this company, and were paid as consultants for their expertise. One of the authors became a salaried employee at Hinge Health following the study, while others remain as clinical advisors who receive equity compensation. These relationships may have introduced bias into this study.
  • A detailed list of exercises and relevant progressions was not shared by the authors, which limits replication of the study.
  • While the authors stated that older participants were more likely to complete the program, while those with higher BMI were less likely complete the program, they were not specific with cut-off thresholds for each category.
  • The authors did not elucidate phenotypic differences between completers who met the pain reduction threshold, and those who do not. Similarly, apart from BMI and age, the authors did not describe the characteristics of non-completers. A detailed description and analysis of why individuals did not complete the program is very clinically important, so as so ‘weed-out’ patients who might not benefit from DCPs in the real-world.
  • Many times, the authors only reported mean changes in measured variables during the study. They did not report outlier scores, median or mode.
  • Lack of control group and tracking of long-term outcomes.
  • This study was conducted through employers and included an extremely large sample size in a real-world setting.
  • The study had similar age and sex distributions for knee and back pain participants, enabling direct comparison between pathways.
  • The study demonstrates an improvement in self-reported workplace productivity (WPAI) and surgery likelihood, suggesting that a DCP might have considerable economic benefits.

Additional References:

  1. Qaseem A, Wilt TJ, McLean RM, Forciea MA, Clinical Guidelines Committee of the American College of Physicians. Noninvasive treatments for acute, subacute, and chronic low back pain: A clinical practice guideline from the American College of Physicians. Ann Intern Med 2017 Apr 4; 166(7): 514–530. PMID: 28192789
  2. Moseley JB, O’Malley K, Petersen NJ, Menke TJ, Brody BA, Kuykendall DH, et al. A controlled trial of arthroscopic surgery for osteoarthritis of the knee. N Engl J Med 2002 Jul 11; 347(2): 81–88. PMID: 12110735
  3. Delitto A, Piva SR, Moore CG, Fritz JM, Wisniewski SR, Josbeno DA, et al. Surgery versus nonsurgical treatment of lumbar spinal stenosis: a randomized trial. Ann Intern Med 2015 Apr 7; 162(7): 465–473. PMID: 2584499.