- Question: Is activity tracker–measured ambulation an accurate and reliable determinant of postoperative quality of recovery, a health status measure?
- Findings: The findings of this prospective cohort validate the clinical use of activity trackers by demonstrating their accuracy as well as their inter- and intradevice reliability for measuring postoperative ambulation; furthermore, ambulation partly explains variations in the quality of recovery.
- Meaning: Ambulation is a determinant of postoperative quality of recovery, and, hypothetically, interventions that improve postoperative ambulation may also enhance quality of recovery; this facilitates using quality of recovery as an outcome measure in studies of anesthetic and analgesic interventions.
Modern health care is progressively adopting enhanced recovery after surgery clinical pathways1 that emphasize early return to function.2 Quality of recovery (QOR) instruments (eg, QoR-15)3 evaluate patient’s ability to resume normal life activities and functionality in the immediate postoperative period. Acute pain is a strong QOR determinant4,5; consequently, QOR measures are gaining popularity in studies of analgesic interventions.6,7 Nonetheless, QOR domains also reflect physical independence.3 Intuitively, interventions that improve perioperative ambulation should also enhance QOR, but this relationship between ambulation and QOR measures has not been explored. Consequently, researchers evaluating anesthetic/analgesic interventions that enhance perioperative ambulation underutilize QOR instruments, failing to capture potential health status benefits.8,9 A recent meta-analysis of regional anesthesia for knee arthroplasty10 illustrates this issue: despite prioritizing motor-sparing strategies and early ambulation,10 QOR was seldom assessed.10,11 This limits our understanding of how similarly effective analgesic interventions may impact health status differently.12
The second impediment to establishing the relationship between ambulation and QOR relates to measuring ambulation. Contemporary care emphasizes early postoperative ambulation,13,14 but quantifying functional recovery employs relatively short heterogeneous tests (eg, 6-minute walk, timed up and go).11 These tests require trained assessors, depend on patients’ comfort/ability to complete the test during a scheduled time window, and may have little bearing on ability to resume normal activities. Self-report ambulation questionnaires are an alternative, but they are subjective, susceptible to recall bias,15 and have modest reliability.16 In contrast, activity trackers are novel, noninvasive, patient-friendly devices that promise objective measurement of ambulation.16,17 Importantly, activity trackers evaluate ambulation during spontaneous activity, over longer time periods, without requiring an assessor. However, while the feasibility of perioperative activity trackers use has been demonstrated,12 clinical utility remains limited by lack of systematic validation of accuracy and reliability in clinical settings.18
We selected a cesarean delivery cohort as a surgical model to address the 2 aforementioned issues. It is a highly reliable, consistent, and generalizable model emphasizing early postoperative ambulation for its effects on maternal recovery, newborn care, thromboembolism prevention, and hospital discharge.19,20 We sought validating activity tracker use for quantifying ambulation as a first objective and exploring the association between ambulation and QOR measures (QoR-15) as a second objective. We hypothesized that (1) activity trackers accurately and reliably measure postoperative ambulation and (2) ambulation is at least moderately correlated with QoR-15 scores in the first 24 hours following cesarean delivery.
This study is reported in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology guidelines.21 This prospective nested cohort study was originally conducted between July 2015 and December 2016 and subsequently extended to June 2017, at St Michael’s Hospital, a tertiary care center affiliated with the University of Toronto (Toronto, ON, Canada). Patients were nested within mode of delivery, as part of a bigger study approved by the hospital’s research ethics board. The bigger study used activity trackers to measure early postpartum ambulation among consented healthy parturients during the first 24 hours following elective uncomplicated vaginal and cesarean deliveries to evaluate an institutional early postpartum ambulation policy.12
This original larger study cohort included nulliparous and uniparous patients having their first and second deliveries, respectively. The present nested cohort included parturients12 who completed an elective lower-segment (Pfannenstiel) cesarean delivery under a spinal anesthetic and who received the standardized institutional multimodal postdelivery pain management protocol. Data relating to postpartum ambulation, analgesia, and QOR during the first 24 hours postdelivery were available from wrist-worn activity trackers that recorded the number of steps taken by each patient and from a patient diary that documented pain and QoR-15 scores. Demographic characteristics (age, body mass index [BMI], fetal weight) were extracted from medical records.
Assessment of QoR-15 Determinants
Participants were interviewed prior to their scheduled delivery to be introduced to the study, familiarized with the wrist-worn activity trackers (UP, Jawbone Fitness Trackers, San Francisco, CA), and acquainted with the patient diary used in registering rest/dynamic pain and QOR (QoR-15) scores.3 Patients were seen again immediately postdelivery to be fitted with the activity trackers and provided with the diary. Trackers were worn continuously around the wrist of the nondominant hand. All trackers were of the same model and used the same software version. Participants were encouraged to ambulate early, as soon as spinal anesthesia wore off (Bromage score = 0). To achieve that, removal of the urinary catheter within 6–8 hours postcesarean, showering within the first 24 hours, and frequent scheduled reminders to ambulate were set as practical milestones. Both trackers and diaries were collected 24 hours following cesarean delivery. The number of steps was retrieved by synchronizing to the Jawbone smartphone application, while the pain scores (2, 6, 12, 18, and 24 hours) and QoR-15 scores (12 and 24 hours) were extracted from the diary. The trackers were then reset and sterilized using a cleaning solution before subsequent use.
Additionally, to determine the patient’s acceptable symptom state (PASS)22,23 with respect to ambulation, patients were requested to rate their opinion of their ambulation at 24 hours postcesarean. Specifically, we asked patients “In your opinion, did you have good ambulation (ie, mobility, walking) in the past 12 hours?” and provided the answer options: “yes, no, not sure.”
Assessment of the Accuracy and Reliability of Activity Trackers
To evaluate the accuracy and reliability of the activity tracker–measured ambulation, a subgroup of participants who completed the 24-hour observation period were involved in an additional assessment. The first 48 patients who approved were asked to walk the corridor in the labor and delivery unit from one end to the other and back (approximately 130 m), at a normal pace—a “corridor walk.” During this walk, each participant was fitted with 2 activity trackers around the same nondominant wrist and was accompanied by a trained assessor who silently counted the patient’s steps. The 2 activity trackers were synchronized to the smartphone application to document the number of steps for each length of the walk.
Study Design and Analysis
To examine the association between postcesarean ambulation during the first 24 hours postdelivery (number of steps during the first 24 hours) and postpartum QOR at 24 hours, we used both regression analysis and Spearman correlation coefficient (ρ). The combination of both regression and correlation analyses allows exploring the strength of the fixed association as well as the dynamic relationship (in response to change). In the regression analysis, we developed a multivariable linear regression model using QoR-15 scores as a dependent variable and postpartum ambulation as the main covariate of interest. The additional covariates considered for inclusion in the model were 9 clinically important explanatory variables (confounders) that were identified a priori as possible determinants of postcesarean QOR, including (1) maternal age, (2) maternal BMI, (3) fetal weight, (4) parity, (5) delivery time (morning versus afternoon cesarean delivery), (6) time to resolution of spinal anesthesia (measured from arrival in recovery room to time when Bromage score was zero), (7) breastfeeding (yes/no), (8) rest pain during the first 24 hours (measured as area under the curve [AUC]), and (9) dynamic pain during the first 24 hours (measured as AUC for pain scores at 2, 6, 12, 18, and 24 hours). The assumptions of linear regression for this model (linear relationship, multivariate normality, minimal or no multicollinearity, no autocorrelation, and homoscedasticity) were verified by visual inspection of data and residuals. For the correlation analysis, the Spearman correlation coefficient was used to quantify the association between each of these factors and QoR-15 score. Correlation was characterized as very weak (ρ = 0.00–0.19), weak (ρ = 0.20–0.39), moderate (ρ = 0.40–0.59), strong (ρ = 0.60–0.79), and very strong (ρ = 0.80–1.00).24 We also separately measured the correlation between each of the 5 domains in the QoR-15 scale with ambulation using the Spearman coefficient.
To determine the PASS value for ambulation following cesarean delivery, we calculated the 75th percentile of ambulation (number of steps) in parturients who rated their ambulation as good at 24 hours postdelivery.22,25,26 Subsequently, we used the previously defined PASS value for QoR-15 in surgical patients,27 a score of 112 on QoR-15, to evaluate how well measurement of ambulation could discriminate between patients who have good QOR and those who do not, using a receiver operating characteristic curve.28
To evaluate the accuracy of activity tracker–measured ambulation, we used the intraclass correlation coefficient (ICC) to quantify the agreement between the total number of steps registered by activity trackers during the “corridor walk” and the actual number of patient steps counted by the accompanying assessor during the same walk. This specific ICC statistic involved a 2-way random-effects model to assess for absolute agreement for a single rater/measurement. ICC values indicate the extent of agreement as follows: 0 to 0.40, poor to fair; 0.41 to 0.60, moderate; 0.61 to 0.80, substantial; and 0.81 to 1.00, almost perfect.29 Agreement between the 2 measurements was further assessed using the Bland–Altman–Tukey technique, which plots the mean differences and the limits of agreement between the 2 measurements of ambulation.30 The Shapiro–Wilk test was used to confirm the normality of data distribution for the difference. Finally, the Lin31 and Lin et al32 concordance correlation coefficient between the assessor-measured and activity tracker–measured ambulation was calculated; and the best fit and the perfect agreement lines were plotted. Precision and accuracy were also described using the Pearson correlation coefficient and the bias correction factor, respectively.
To assess the reliability of activity trackers, we evaluated both inter- and intradevice reliability using the ICC coefficient, as described by Shrout and Fleiss.29 For interdevice reliability, we examined the correlation between the total number of steps registered by each of the 2 activity trackers worn around the wrist of the same patient during the “corridor walk.” This specific ICC statistic involved a 2-way random-effects model to assess for absolute agreement between 2 (or more) raters. For intradevice reliability, we examined the agreement between the numbers of steps registered by an activity tracker during each length of the “corridor walk.” This specific ICC statistic involves a 2-way random-effects model to assess for absolute agreement between 2 (or more) measurements of the same rater. The ICC statistic is ideal for reliability testing as it accounts for observer bias, agreement attributed to chance, and also has the capacity to capture actual concordance rather than trends.33
We designated 0.05 as a threshold of statistical significance for the 2-tailed testing of the associations investigated. The Bonferroni–Holm correction was used to adjust for repeated measurements of QoR-15 scores.34 Sequential hypothesis testing permitted maintaining the threshold of statistical significance at 0.05 for the hypotheses tested.35 Continuous outcomes (eg, ambulation) were reported as mean (95% confidence interval [CI]), and ordinal outcomes (eg, QoR-15 and pain visual analogue scale scores) were treated as continuous data. Categorical outcomes were reported as proportions (percentages). Continuous outcomes were compared using the Mann–Whitney U test or Student t test, and categorical outcomes were compared using the Fisher exact test or χ2 test. We analyzed data using R statistical package version 3.4.2 (R Foundation for Statistical Computing, Vienna, Austria).
Sample Size Calculation
We sequentially tested 2 hypotheses36 using a serial gate-keeping approach,35 beginning with the accuracy and reliability of ambulation measurement and subsequently moving to the correlation between ambulation and QoR-15 scores. Such serial testing is clinically justified in the specific setting of our cohort study because we were interested in evaluating the correlation between ambulation and QoR-15 scores only if activity trackers themselves provided accurate and reliable measurements.
To test the hypothesis that ambulation was a determinant of postoperative QOR, we used a multivariable linear regression model that included the preidentified potential covariates. Sample size calculation was performed under the assumptions of linearity, normality of residuals distribution, and independence of the variables examined. For a probability level (α) equivalent to .05 and an anticipated effect size equivalent to 0.15, we estimated that a sample size of 118 patients would provide 80% power for the multiple regression study.37 To account for any potential correlation between the independent variables,38 we decided to inflate the sample size to 200 patients.
To test the hypothesis that the activity tracker–measured ambulation in an accurate and reliable manner, we used the ICC coefficient. Our earlier work12 suggested that the ICC for the agreement between assessor-measured and activity tracker–measured ambulation is “almost perfect” (ie, ICC = 0.8). To estimate the sample size for the ICC testing, we used the graphs presented by Eliasziw et al.39 We confirmed the adequacy of this estimate using the data outlined by Zou.40 With an α value of .05, we estimated that 48 patients would provide 80% power for the planned ICC test. For reliability testing using the ICC, we aimed to demonstrate an “almost perfect” intradevice reliability (ie, ICC = 0.8) and a “substantial” interdevice reliability (ie, ICC = 0.6). With 2 devices measuring ambulation per patient, 2 measurements per device, and an α of .05, we estimated that 40 patients would provide 80% power for the planned correlation testing.
Based on the above sample size calculations for this prospective nested cohort, we planned to include 200 patients in the study evaluating whether ambulation is a QOR determinant. Of those, the first consecutive 48 patients who approved the “corridor walk” were included in the study of accuracy and reliability of activity trackers.
Two hundred consecutive consented cesarean delivery patients were included in this nested cohort, of which 48 also completed the accuracy and reliability component of the study. Of these 200, only 23 patients had been included in our previously published feasibility study.12 All patients had a low-risk pregnancy followed by an uncomplicated cesarean delivery performed during regular work hours (8:00–16:00) between July 2015 and June 2017. Data were complete for all patients and were included in the analysis. Table 1 summarizes the demographic characteristics and the ambulation as well as pain severity and QoR-15 scores of the cohort. Expressed as a mean (standard deviation), the participants took a total of 783 (967) steps and reported a QoR-15 score of 124.7 (28.9) by the 24-hour time point following cesarean delivery.
Assessment of Activity Tracker Accuracy and Reliability
Comparing the device- to the assessor-measured ambulation revealed that these 2 measurements were almost in perfect agreement, with an ICC (95% CI) of 0.93 (0.91–0.95), suggesting that activity trackers are capable of accurately measuring ambulation. Evaluation of the inter- and intradevice reliability yielded ICC (95% CI) values of 0.98 (0.97–0.99) and 0.96 (0.93–0.99), respectively, suggesting excellent reliability (Table 2).
Additionally, assessment of the extent of agreement between device- and assessor-measured ambulation using the Bland–Altman plot revealed a mean observed difference (95% limits of agreement) equivalent to −2.6 steps (−3.7 to 9.0; Figure, panel A). A total of 45 of 48 parturients (94%) fell within a 13-step range, from the 171 steps walked during the “corridor walk.” Moreover, assessment of the agreement between the assessor- and the device-measured ambulation using the concordance coefficient revealed almost perfect concordance (0.994), with high accuracy (0.996), and high precision (0.992; Figure, panel B).
Assessment of QOR Determinants
Among the preidentified potential covariates, multivariable linear regression modeling found that maternal age, fetal weight, and time to spinal anesthesia resolution were not significantly associated with QoR-15 score following cesarean delivery (Table 3). In contrast, the remaining 7 covariates (ambulation, maternal BMI, 24-hour rest pain AUC, 24-hour dynamic pain AUC, parity, delivery time, and breastfeeding) were significantly associated with QoR-15 score. Among these covariates, the association between ambulation and QoR-15 scores following cesarean delivery was the strongest. The linear regression parameter estimate (95% CI) for ambulation was 0.002 (0.001–0.003). Furthermore, the Spearman correlation coefficient (ρ) (95% CI) for ambulation was 0.560 (0.328–0.728), suggesting moderate correlation. The linear regression and Spearman correlation results for the remaining covariates are presented in Table 3.
The overall coefficient of determination (R2) for linear regression after accounting for the 7 covariates was 0.753, while the partial coefficients of determination for ambulation and dynamic pain were 0.311 and 0.135, respectively; thus, ambulation and dynamic pain were the main covariates in this model that explained the variability observed in the QoR-15 score (Table 3).
When this association was further explored according to the domains of the QoR-15 scale, Spearman correlation analysis suggested that ambulation was not associated with psychological support (Table 4). In contrast, ambulation was strongly associated with physical independence (ρ = 0.79 [0.731–0.837], P = .0008) and physical comfort (ρ = 0.64 [0.550–0.715], P = .006), moderately associated with emotional state (ρ = 0.58 [0.480–0.665], P = .01), and weakly associated with postoperative dynamic pain (ρ = 0.04 [0.01–0.178], P = .04).
PASS for Ambulation and Its Explanatory Value
The 75th percentile for the number of steps for parturients who reported good ambulation was 287 steps, reflecting the cutoff above which patients feel that they had good ambulation in the first 24 hours following cesarean delivery (Table 5). With a previously estimated PASS value for QoR-15 score at 24 hours equivalent to 112,27 the area under the receiver operating characteristic curve was estimated to be 0.84 (95% CI, 0.76–0.92). This estimate confirms that activity tracker–measured ambulation possesses good discriminatory ability between parturients who have good QOR following cesarean delivery and those who do not.
This nested cohort study demonstrated the accuracy of activity trackers, as well as their inter- and intradevice reliability, for measuring postoperative ambulation in the clinical settings examined. Furthermore, exploring the relationship between ambulation and QOR following cesarean delivery revealed that ambulation, on its own, partly explained variations in QoR-15 overall score, as well as the scores of its individual domains, apart from psychological support. When combined with dynamic pain, ambulation explained nearly half of variation in QoR-15 scores. Furthermore, a total of 287 steps were found to represent the “patients’ acceptable symptom status” value for postoperative ambulation, beyond which patients considered that they had good ambulation during the first 24 hours following cesarean delivery.
Our findings carry the hypothetical implication that interventions aimed at improving ambulation may also help enhance QOR, that is, health status, in surgical patients. The importance of these findings is best understood in terms of facilitating the integration of health status measures, such as QOR instruments, into the core of outcomes measured when anesthetic and/or analgesic interventions that influence postoperative ambulation are investigated. Furthermore, as early ambulation is a common theme in enhanced recovery after surgery41 and prevention of deep vein thrombosis recommendations,42 our results provide a novel, validated, and objective tool to quantify postsurgical ambulation, enabling assessment of strategies to enhance postoperative patient mobilization. While the population examined herein received neuraxial anesthesia, we believe that the evidence is easily transferable to a wide range of surgical populations, with emerging use as an indicator of lifestyle changes following thoracic and orthopedic surgery.43,44 Researchers can now focus on prospectively examining interventions that improve postoperative ambulation and QOR. For example, lower extremity surgical procedures that prioritize postoperative ambulation and utilize motor-sparing nerve blocks are an obvious candidate population. Finally, PASS value we calculated can be used to power future studies and identify responders in studies examining interventions that improve ambulation.
The current study is novel in ascertaining the relationship between ambulation and QOR measures. Our previous study12 detected a possible signal when exploring potential correlations between QoR-15 score on the one hand, and several measured outcomes, including rest pain, dynamic pain, and ambulation. However, the post hoc nature of the analysis and lack of sufficient statistical power undermined the validity of this finding. The QoR-15 scale3 incorporates items that rate the ability to return to activity and to look after personal toilet and hygiene unaided, as well as items rating frequency of experiencing moderate and severe pain. However, ambulation and pain severity during movement (dynamic) per se are neither directly measured by nor are they part of the QoR-15 construct. In addition, our work is unique in providing validation of activity trackers for measurement of early postoperative ambulation in clinical settings. Existing validation studies have focused on other aspects, such as validating the activity tracker measurement of energy expenditure,45,46 sleep time,18 distance covered,47 and interdevice reliability for different brands.18 The paucity of studies examining intradevice reliability (same brand) is also noted.18
Our validation study estimated the mean difference between the real (assessor measured) and estimated (device measured) ambulation to be 2.6 steps (1.5%), with a range equivalent to 7.6% of the number of steps taken. This magnitude of error is considered acceptable, commensurate with other observations,48 and supports the clinical utility of these devices. However, the fact that the best fit line was noted to be almost parallel and above the perfect agreement line may suggest a small yet consistent systematic error in estimating the number of steps by activity tracker devices. This observation is not new because other studies have also noted the tendency of activity trackers to overestimate the number of steps.18,49,50 In fact, our understanding of the limitations of accelerometer algorithms used in activity trackers has been advanced by various studies showing that wrist-worn trackers generally tend to slightly overestimate ambulation because they interpret upper extremity movement as steps,51 while chest- and waist-worn trackers seem to have higher accuracy.48,52 Furthermore, accuracy of these devices is also reduced when measuring ambulation at slower speeds,48,53 or with the help of a walking aid.52 That said, the specific device used in this study (UP by Jawbone) is reported to be less prone to overestimate (by interpreting upper extremity movement as ambulation)54 and to retain its accuracy at slow speeds.48 Notwithstanding, the enhanced accuracy of the specific device we used may itself restrict the generalizability of our conclusions to other activity trackers.
Our study has several limitations. First, its design, a nested prospective cohort, allows informing associations, but not causal relationships or their respective directions. Second, the population from which this nested cohort was derived was limited to low-risk pregnancies having uncomplicated cesarean deliveries occurring during daytime hours and observed for 24 hours postpartum. Also, the corridor walk used in the validation may not accurately reflect the complexity of the postoperative clinical environment. Additionally, commercial activity trackers use nonstandardized algorithms that may be updated without notice, limiting the ability to conduct comparison across models or even across different periods of time for the same model. Thus, our findings may not be generalizable to different clinical settings, cultures, or centers with more constrained postpartum ambulation policies. Third, the a priori choice of covariates for analysis was governed by the authors’ judgment of clinical relevance as well as availability of data. For example, opioid requirements were not included, as our previous study showed very limited postoperative opioid consumption in this population. Additionally, data were not available for some potentially relevant covariates, such as surgical blood loss, time to discontinuing intravenous infusion and Foley catheter, heaviness of lochia, postpartum dizziness, sedation or nausea/vomiting, newborn health status, presence of an assistant (spouse or family member), and sleep deprivation. Fourth, quantifying ambulation was conducted in real clinical rather than controlled settings. Therefore, we cannot exclude the possibility that human factors may have influenced the step length46,51 during the “corridor walk” exercise. For example, distraction (parturient or assessor), pain, fatigue, and abdominal/pelvic cramps may all interfere with step length. Fifth, the use of activity trackers, on their own, seems to influence behavioral changes toward reduced sitting and bed rest and increased ambulation.55,56 Sixth, we relied on step counting as a benchmark, which is an acceptable approach used in such validation studies53; privacy concerns preclude using video recording, which is arguably more accurate.48,57 Seventh, while we inflated the sample size (by 67%) to account for potential relationships between covariates, the adequate power for multivariable regression analysis remains a debated topic.38 Finally, our study was limited to 1 manufacturer brand; we did not examine interdevice reliability for different brands.
In summary, our observational study demonstrates the accuracy and reliability of using activity trackers to measure ambulation in the postsurgical clinical settings. The study also suggests that activity tracker–measured ambulation is a determinant of postoperative QOR. This begs the hypothesis that interventions aimed at improving ambulation may help also enhance QOR in surgical patients. These findings encourage using QOR instruments to measure improvements in postoperative ambulation and investigating the mechanisms by which interventions can impact QOR.
The authors are grateful for the valuable help received from research volunteers with the Departments of Anesthesia and Obstetrics and Gynecology, St Michael’s Hospital, University of Toronto, Toronto, ON, Canada. Among others, we wish to thank Khaled Soliman, Nilab Mirzada, Rakhi Tilak, and Samir Shamji for their effort and time. We are also thankful to Dr Colin J. L. McCartney for his thoughtful suggestions.
Name: Faraj Massouh, MD.
Contribution: This author helped conduct the study, analyze the data, and prepare the manuscript.
Name: Rachel Martin, MD, FRCPC.
Contribution: This author helped conduct the study and prepare the manuscript.
Name: Bokman Chan, MD, FRCPC.
Contribution: This author helped conduct the study and prepare the manuscript.
Name: Julia Ma, MPH.
Contribution: This author helped conduct the study and prepare the manuscript.
Name: Vikita Patel, BSc.
Contribution: This author conduct the study and prepare the manuscript.
Name: Michael P. Geary, MB, BCh, BAO, FRCOG, FRCPI, MD, DCH.
Contribution: This author helped design the study and prepare the manuscript.
Name: John G. Laffey, MD, MA, FCAI.
Contribution: This author helped with study design and manuscript preparation.
Name: Duminda N. Wijeysundera, MD, PhD, FRCPC.
Contribution: This author helped analyze the data and prepare the manuscript.
Name: Faraj W. Abdallah, MD.
Contribution: This author helped conceive the study idea, design the study, conduct the study, analyze the data, and prepare the manuscript.
This manuscript was handled by: Richard Brull, MD, FRCPC.
1. Melnyk M, Casey RG, Black P, Koupparis AJ. Enhanced recovery after surgery (ERAS) protocols: time to change practice? Can Urol Assoc J. 2011;5:342–348.
2. Fearon KC, Ljungqvist O, Von Meyenfeldt M, et al. Enhanced recovery after surgery: a consensus review of clinical care for patients undergoing colonic resection. Clin Nutr. 2005;24:466–477.
3. Stark PA, Myles PS, Burke JA. Development and psychometric evaluation of a postoperative quality of recovery score: the QoR-15. Anesthesiology. 2013;118:1332–1340.
4. Wu CL, Naqibuddin M, Rowlingson AJ, Lietman SA, Jermyn RM, Fleisher LA. The effect of pain on health-related quality of life in the immediate postoperative period. Anesth Analg. 2003;97:1078–1085.
5. Wu CL, Rowlingson AJ, Partin AW, et al. Correlation of postoperative pain to quality of recovery in the immediate postoperative period. Reg Anesth Pain Med. 2005;30:516–522.
6. Abdallah FW, Morgan PJ, Cil T, et al. Ultrasound-guided multilevel paravertebral blocks and total intravenous anesthesia improve the quality of recovery after ambulatory breast tumor resection. Anesthesiology. 2014;120:703–713.
7. Elkassabany N, Cai LF, Mehta S, et al. Does regional anesthesia improve the quality of postoperative pain management and the quality of recovery in patients undergoing operative repair of tibia and ankle fractures? J Orthop Trauma. 2015;29:404–409.
8. Herrera FJ, Wong J, Chung F. A systematic review of postoperative recovery outcomes measurements after ambulatory surgery. Anesth Analg. 2007;105:63–69.
9. Gornall BF, Myles PS, Smith CL, et al. Measurement of quality of recovery using the QoR-40: a quantitative systematic review. Br J Anaesth. 2013;111:161–169.
10. Terkawi AS, Mavridis D, Sessler DI, et al. Pain management modalities after total knee arthroplasty: a network meta-analysis of 170 randomized controlled trials. Anesthesiology. 2017;126:923–937.
11. Choi S, Trang A, McCartney CJ. Reporting functional outcome after knee arthroplasty and regional anesthesia: a methodological primer. Reg Anesth Pain Med. 2013;38:340–349.
12. Ma J, Martin R, Chan B, et al. Using activity trackers to quantify postpartum ambulation: a prospective observational study of ambulation after regional anesthesia and analgesia interventions. Anesthesiology. 2018;128:598–608.
13. McCartney CJ, Wong P. How can we best balance pain control and rehabilitation after knee replacement? Anesth Analg. 2016;122:1760–1762.
14. Ilfeld BM, McCartney CJL. Searching for the optimal pain management technique after knee arthroplasty: analgesia is just the tip of the iceberg. Anesthesiology. 2017;126:768–770.
15. Castelino T, Fiore JF Jr, Niculiseanu P, Landry T, Augustin B, Feldman LS. The effect of early mobilization protocols on postoperative outcomes following abdominal and thoracic surgery: a systematic review. Surgery. 2016;159:991–1003.
16. Berlin JE, Storti KL, Brach JS. Using activity monitors to measure physical activity in free-living conditions. Phys Ther. 2006;86:1137–1145.
17. Appelboom G, Yang AH, Christophe BR, et al. The promise of wearable activity sensors to define patient recovery. J Clin Neurosci. 2014;21:1089–1093.
18. Evenson KR, Goto MM, Furberg RD. Systematic review of the validity and reliability of consumer-wearable activity trackers. Int J Behav Nutr Phys Act. 2015;12:159.
19. Karlström A, Engström-Olofsson R, Norbergh KG, Sjöling M, Hildingsson I. Postoperative pain after cesarean birth affects breastfeeding and infant care. J Obstet Gynecol Neonatal Nurs. 2007;36:430–440.
20. Say L, Chou D, Gemmill A, et al. Global causes of maternal death: a WHO systematic analysis. Lancet Glob Health. 2014;2:e323–e333.
21. von Elm E, Altman DG, Egger M, Pocock SJ, Gøtzsche PC, Vandenbroucke JP; STROBE Initiative. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies. Prev Med. 2007;45:247–251.
22. Tubach F, Wells GA, Ravaud P, Dougados M. Minimal clinically important difference, low disease activity state, and patient acceptable symptom state: methodological issues. J Rheumatol. 2005;32:2025–2029.
23. Wariaghli G, Allali F, Berrada K, et al. The patient acceptable symptom state of chronic musculoskeletal pain measured on a visual analog scale in Moroccan patients. Pain Med. 2013;14:103–109.
24. Spearman C. The proof and measurement of association between two things. Am J Psychol. 1904;15:72–101.
25. Kvien TK, Heiberg T, Hagen KB. Minimal clinically important improvement/difference (MCII/MCID) and patient acceptable symptom state (PASS): what do these concepts mean? Ann Rheum Dis. 2007;66(suppl 3):iii40–iii41.
26. Tubach F, Ravaud P, Beaton D, et al. Minimal clinically important improvement and patient acceptable symptom state for subjective outcome measures in rheumatic disorders. J Rheumatol. 2007;34:1188–1193.
27. Myles PS, Myles DB, Galagher W, Chew C, MacDonald N, Dennis A. Minimal clinically important difference for three quality of recovery scales. Anesthesiology. 2016;125:39–45.
28. Sedgwick P. How to read a receiver operating characteristic curve. BMJ. 2015;350:1–3.
29. Shrout PE, Fleiss JL. Intraclass correlations: uses in assessing rater reliability. Psychol Bull. 1979;86:420–428.
30. Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet. 1986;1:307–310.
31. Lin LI. A concordance correlation coefficient to evaluate reproducibility. Biometrics. 1989;45:255–68.
32. Lin L, Hedayat AS, Sinha B, Yang M. Statistical methods in assessing agreement: models, issues, and tools. J Am Stat Assoc. 2002;97:257–270.
33. Kramer MS, Feinstein AR. Clinical biostatistics. LIV. The biostatistics of concordance. Clin Pharmacol Ther. 1981;29:111–123.
34. Holm S. A simple sequentially rejective multiple test procedure. Scan J Statist. 1979;6:65–70.
35. Mascha EJ, Turan A. Joint hypothesis testing and gatekeeping procedures for studies with multiple endpoints. Anesth Analg. 2012;114:1304–1317.
36. Dmitrienko A, Tamhane AC. Gatekeeping procedures with clinical trial applications. Pharm Stat. 2007;6:171–180.
37. Green SB. How many subjects does it take to do a regression analysis. Multivariate Behav Res. 1991;26:499–510.
38. Maxwell SE. Sample size and multiple regression analysis. Psychol Methods. 2000;5:434–458.
39. Eliasziw M, Young SL, Woodbury MG, Fryday-Field K. Statistical methodology for the concurrent assessment of interrater and intrarater reliability: using goniometric measurements as an example. Phys Ther. 1994;74:777–788.
40. Zou GY. Sample size formulas for estimating intraclass correlation coefficients with precision and assurance. Stat Med. 2012;31:3972–3981.
41. Stethen TW, McLoughlin JM, Heidel E, Barnes L. Walking to recovery: the effects of missed ambulation on postsurgical recovery. J Am Coll Surg. 2016;223:e116.
42. Nicolaides A, Fareed J, Kakkar AK, et al. Prevention and treatment of venous thromboembolism: International Consensus Statement (guidelines according to scientific evidence). Int Angiol. 2013;19:116–118.
43. Cook DJ, Thompson JE, Prinsen SK, Dearani JA, Deschamps C. Functional recovery in the elderly after major surgery: assessment of mobility recovery using wireless technology. Ann Thorac Surg. 2013;96:1057–1061.
44. Toogood PA, Abdel MP, Spear JA, Cook SM, Cook DJ, Taunton MJ. The monitoring of activity at home after total hip arthroplasty. Bone Joint J. 2016;98-B:1450–1454.
45. Price K, Bird SR, Lythgo N, Raj IS, Wong JY, Lynch C. Validation of the Fitbit One, Garmin Vivofit and Jawbone UP activity tracker in estimation of energy expenditure during treadmill walking and running. J Med Eng Technol. 2017;41:208–215.
46. Sasaki JE, Hickey A, Mavilia M, et al. Validation of the Fitbit wireless activity tracker for prediction of energy expenditure. J Phys Act Health. 2015;12:149–154.
47. Wahl Y, Düking P, Droszez A, Wahl P, Mester J. Criterion-validity of commercially available physical activity tracker to estimate step count, covered distance and energy expenditure during sports conditions. Front Physiol. 2017;8:725.
48. Chow JJ, Thom JM, Wewege MA, Ward RE, Parmenter BJ. Accuracy of step count measured by physical activity monitors: the effect of gait speed and anatomical placement site. Gait Posture. 2017;57:199–203.
49. Van Blarigan EL, Kenfield SA, Tantum L, Cadmus-Bertram LA, Carroll PR, Chan JM. The Fitbit One physical activity tracker in men with prostate cancer: validation study. JMIR Cancer. 2017;3:e5.
50. Alharbi M, Bauman A, Neubeck L, Gallagher R. Validation of Fitbit-Flex as a measure of free-living physical activity in a community-based phase III cardiac rehabilitation population. Eur J Prev Cardiol. 2016;23:1476–1485.
51. Petrucci GJ, Masterller BR, Cox MF, Staudenmayer JW, Sirard JR, Freedson PS. A consumer activity tracker is sensitive to changes in steps during simulated free-living. Med Sci Sports Exerc. 2017;49(5S):493.
52. Alinia P, Cain C, Fallahzadeh R, Shahrokni A, Cook D, Ghasemzadeh H. How accurate is your activity tracker? A comparative study of step counts in low-intensity physical activities. JMIR Mhealth Uhealth. 2017;5:e106.
53. Wong CK, Mentis HM, Kuber R. The bit doesn’t fit: evaluation of a commercial activity-tracker at slower walking speeds. Gait Posture. 2018;59:177–181.
54. Storm FA, Heller BW, Mazzà C. Step detection and activity recognition accuracy of seven physical activity monitors. PLoS One. 2015;10:e0118723.
55. Gualtieri L, Rosenbluth S, Phillips J. Can a free wearable activity tracker change behavior? The impact of trackers on adults in a physician-led wellness group. JMIR Res Protoc. 2016;5:e237.
56. Larouche ML, Mullane SL, Toledo MJ, et al. Objectively measured e-newsletter engagement and workplace sedentary behavior. Annals of Behavioral Medicine. 2017:New York, NY: Springer, S932–S933.
57. Smith M, Powers M, Henderson O, Bodman T. Accuracy and reliability of the FitBit ChargeTM activity tracker among older adults. Oklahoma AHPERD. 2017;54:29–37.