Data on the 0.4-km walk were missing in 8 participants. ADL indicates activity of daily living; SILVER-AMI, Comprehensive Evaluation of Risk Factors in Older Patients With Acute Myocardial Infarction.
ADL indicates activity of daily living.
Adjusted for age, sex, race/ethnicity, educational level, marital status, AMI type, peak troponin ratio, comorbidities, smoking history, in-hospital procedures, in-hospital complications, discharge location, preadmission functional impairment, hearing impairment, visual impairment, grip strength impairment, cognitive impairment, depression, and social support. ADL indicates activity of daily living.
eTable 1A. Multivariable-Adjusted Odds Ratios for Risk of ADL Decline After AMI
eTable 1B. Multivariable-Adjusted Odds Ratios for Risk of Decline in ¼-Mile Mobility After AMI
eTable 2. Multivariable-Adjusted Associations of Mobility Status With Count of Activities in Which Participants Decline at 6 Months After AMI
eFigure 1. Sensitivity Analyses of Associations Between Mobility Impairment and Risk of Functional Decline, Accounting for Competing Risk of Death
eFigure 2. Multivariable-Adjusted Odds Ratios for Risk of Decline in Independence in Each Activity of Daily Living After AMI, According to Mobility Status
Customize your JAMA Network experience by selecting one or more topics from the list below.
Hajduk AM, Murphy TE, Geda ME, et al. Association Between Mobility Measured During Hospitalization and Functional Outcomes in Older Adults With Acute Myocardial Infarction in the SILVER-AMI Study. JAMA Intern Med. 2019;179(12):1669–1677. doi:10.1001/jamainternmed.2019.4114
Is mobility impairment, measured during hospitalization, a “geriatric biomarker” for functional decline among older adults with acute myocardial infarction?
In this cohort study of 2587 patients 75 years or older hospitalized for acute myocardial infarction, mobility impairment measured by the Timed “Up and Go” was significantly associated in a dose-response manner with risk of functional decline.
This study’s findings suggest that a brief and easy-to-administer mobility assessment may be useful in the inpatient setting to identify older patients with acute myocardial infarction who are at risk for functional decline.
Many older survivors of acute myocardial infarction (AMI) experience functional decline, an outcome of primary importance to older adults. Mobility impairment has been proposed as a risk factor for functional decline but has not been evaluated to date in older patients hospitalized for AMI.
To examine the association of mobility impairment, measured during hospitalization, as a risk marker for functional decline among older patients with AMI.
Design, Setting, and Participants
Prospective cohort study among 94 academic and community hospitals in the United States. Participants were 2587 hospitalized patients with AMI who were 75 years or older. The study dates were January 2013 to June 2017.
Main Outcomes and Measures
Mobility was evaluated during AMI hospitalization using the Timed “Up and Go,” with scores categorized as preserved mobility (≤15 seconds to complete), mild impairment (>15 to ≤25 seconds to complete), moderate impairment (>25 seconds to complete), and severe impairment (unable to complete). Self-reported function in activities of daily living (ADLs) (bathing, dressing, transferring, and walking around the home) and walking 0.4 km (one-quarter mile) was assessed at baseline and 6 months after discharge. The primary outcomes were worsening of 1 or more ADLs and loss of ability to walk 0.4 km from baseline to 6 months after discharge. The association between mobility impairment and risk of functional decline was evaluated with multivariable-adjusted logistic regression.
Among 2587 hospitalized patients with AMI, the mean (SD) age was 81.4 (4.8) years, and 1462 (56.5%) were male. More than half of the cohort exhibited mobility impairment during AMI hospitalization (21.8% [564 of 2587] had mild impairment, 16.0% [414 of 2587] had moderate impairment, and 15.2% [391 of 2587] had severe impairment); 12.8% (332 of 2587) reported ADL decline, and 16.7% (431 of 2587) reported decline in 0.4-km mobility. Only 3.8% (30 of 800) of participants with preserved mobility experienced any ADL decline compared with 6.9% (39 of 564) of participants with mild impairment (adjusted odds ratio [aOR], 1.24; 95% CI, 0.74-2.09), 18.6% (77 of 414) of participants with moderate impairment (aOR, 2.67; 95% CI, 1.67-4.27), and 34.7% (136 of 391) of participants with severe impairment (aOR, 5.45; 95% CI, 3.29-9.01). Eleven percent (90 of 800) of participants with preserved mobility declined in ability to walk 0.4 km compared with 15.2% (85 of 558) of participants with mild impairment (aOR, 1.51; 95% CI, 1.04-2.20), 19.0% (78 of 411) of participants with moderate impairment (aOR, 2.03; 95% CI, 1.37-3.02), and 24.6% (95 of 386) of participants with severe impairment (aOR, 3.25; 95% CI, 2.02-5.23).
Conclusions and Relevance
This study’s findings suggest that mobility impairment assessed during hospitalization may be a potent risk marker for functional decline in older survivors of AMI. These findings also suggest that brief, validated assessments of mobility should be part of the care of older hospitalized patients with AMI to identify those at risk for this important patient-centered outcome.
Acute myocardial infarction (AMI) is increasingly a condition of older adults, with more than one-third of AMIs occurring in patients 75 years or older.1 Owing to increased use of therapies such as percutaneous coronary intervention among older adults,2 survival after AMI in this age group has improved.3,4 However, many older survivors of AMI face another major challenge to their health, namely, functional decline (ie, loss of ability to independently perform everyday activities). Thirty percent of survivors of AMI report functional decline after AMI,5-7 and AMI is recognized as a common contributor to the development of disability in older adults.6-10 Functional decline, in addition to directly diminishing independence and quality of life,11 is a harbinger of poor outcomes, such as readmissions,12 institutionalization,13 and mortality.14
Maintenance of function is an outcome valued most by many older adults.15 Despite this, little is known about factors that put older adults at risk for functional decline after hospitalization for AMI. One common vulnerability experienced by many older hospitalized adults is impairment in mobility (ie, difficulty with transfers and ambulation).16 Research has linked poor mobility to increased risk of mortality,17 frailty,18 and functional decline,19 and emerging evidence suggests that mobility impairment is prevalent and predictive of poor outcomes among older patients with AMI.5,20 However, prior studies have been constrained by limited generalizability (eg, single site) or untimely collection of mobility data (ie, after hospitalization). Therefore, we explored the association between mobility, assessed during AMI hospitalization, and risk of functional decline at 6 months after discharge in a prospective nationwide cohort of adults 75 years or older.
Data were obtained from the Comprehensive Evaluation of Risk Factors in Older Patients With Acute Myocardial Infarction (SILVER-AMI) study, a prospective cohort study of 3041 adults 75 years or older hospitalized with AMI (ClinicalTrials.gov identifier R01HL115295). Details of the study have been published previously.21 The study dates were January 2013 to June 2017. Briefly, participants were recruited from 94 academic and community hospitals across the United States. Site coordinators reviewed hospital admission records daily to identify potentially eligible participants and performed medical record review to confirm AMI diagnosis in accord with the Third Universal Definition of Myocardial Infarction.22 Site coordinators approached eligible patients, explained the scope of the study, and obtained written informed consent. The University of California, San Diego Brief Assessment of Capacity to Consent23 was administered to patients with decisional capacity concerns, and proxy consent was obtained for patients with diminished capacity. Patients were ineligible if they had initial troponin elevation more than 24 hours after admission, were transferred from another hospital after more than 24 hours, were incarcerated, or were unable to provide informed consent, with no proxy available. All study protocols were approved by the institutional review boards at the coordinating site (Yale School of Medicine, New Haven, Connecticut) and all participating sites.
Participants underwent a structured interview and physical assessment during AMI hospitalization. Information was collected on geriatric impairments (mobility, hearing, vision, cognition, etc), functional status, demographics, and psychosocial characteristics.
Mobility status was assessed during the baseline interview (median day of admission, day 2; interquartile range, days 2-4) using the Timed “Up and Go” (TUG),24 a performance-based assessment with demonstrated validity in populations with cardiac concerns.25 To complete the TUG, participants were instructed to rise from a seated position, walk 3 m, turn, walk 3 m back to the chair, and sit down. Participants could use assistive devices but were not allowed assistance from another person. We used the time it took participants to complete the maneuver (in seconds) as the score on the TUG. We applied cut points (based on historical TUG scores in hospitalized older adults26,27 and adapted from prior studies28,29) to classify participants into the following groups: preserved mobility (≤15 seconds to complete), mild impairment (>15 to ≤25 seconds to complete), moderate impairment (>25 seconds to complete), and severe impairment (unable to complete assessment, verified via medical record review and communication with site coordinators). Of note, these thresholds differ from those commonly used for the TUG (eg, <10 to 15 seconds),24,30,31 which were established in samples of community-dwelling older adults and may be too conservative for characterizing mobility impairment in acutely ill hospitalized older patients. Missing TUG scores were multiply imputed (by chained equations) for the 16.2% (418 of 2587) of participants who were able to complete the assessment (based on documentation in the medical record review and communication with site coordinators) but did not (eg, refused owing to fear of falling or fatigue).
Demographic characteristics (race/ethnicity, educational level, and marital status) were assessed via self-report during the baseline interview or via medical record abstraction. Geriatric impairments (hearing,32 vision,33 grip strength,34 and cognition35) and psychosocial characteristics (depressive symptoms36 and social support37) were assessed via self-report or performance-based testing during the baseline interview.
Coordinators at each enrollment site and a research nurse at the coordinating center abstracted data from participants’ medical records on presenting clinical characteristics, medical history, procedures (cardiac catheterization only, percutaneous coronary intervention, and coronary artery bypass graft), and complications (arrhythmia, heart failure, bleeding event, and acute kidney injury). Follow-up interviews were conducted with all participants at 6 months after discharge. A physician panel (J.A.D. and S.I.C. and other nonauthors) confirmed deaths between hospital discharge and the 6-month interview via review of medical records and death certificates.
Primary outcomes were defined as a decrease in ability to independently perform 1 or more essential activities of daily living (ADLs) and loss of ability to walk 0.4 km (one-quarter mile) at 6 months after hospital discharge relative to premorbid functional status. During the baseline interview (participants reported on function 1 month before admission) and the 6-month follow-up interview, participants were asked how much help they needed from another person to bathe, dress, transfer (get in and out of a chair), and walk around their home.38 Response options were “no help,” “help,” and “unable to do.” Decline in ADLs was characterized as any decrease in ability to perform these tasks from baseline to 6 months after discharge (ie, transition from “no help” to “help,” “help” to “unable to do,” or “no help” to “unable to do”). Participants were also asked about their ability to walk 0.4 km (or 2-3 blocks), an established marker of neighborhood-level mobility,39 at baseline (reporting on 1 month before admission) and 6 months after discharge. Decline in ability to walk 0.4 km was defined as reporting yes at baseline and no at 6 months after discharge. Declines in individual ADLs were examined as secondary outcomes, as was a count outcome of the number of functional activities in which participants declined (range, 0-5) from baseline to 6 months after discharge.
Baseline characteristics were compared according to level of mobility impairment (ie, preserved mobility [≤15 seconds to complete], mild impairment [>15 to ≤25 seconds to complete], moderate impairment [>25 seconds to complete], or severe impairment [unable to complete]) using analysis of variance for continuous variables, Kruskal-Wallis test for nonnormal/ordinal variables, and χ2 test for categorical variables. Multivariable-adjusted logistic regression was used to model odds of functional decline associated with degree of mobility impairment, with preserved mobility used as the reference category. Negative binomial regression was used to examine the association of mobility impairment with the count of functional activities in which participants declined. Selection of covariates for inclusion in multivariable-adjusted models was guided by the literature,5,40,41 review of bivariate associations between covariates and the outcomes, and clinical judgment and included the following: demographics (age, sex, race/ethnicity, educational level, marital status, and cohabitation status), AMI characteristics (AMI type, peak troponin ratio, and in-hospital revascularization), comorbidities (arrhythmia, heart failure, hypertension, peripheral vascular disease, stroke, prior AMI, chronic obstructive pulmonary disease, cancer, smoking history, Charlson Comorbidity Index, prior revascularization, and chronic lung disease), in-hospital complications (arrhythmia, heart failure, bleeding event, and acute kidney injury), discharge location, geriatric impairments (preadmission impairment in ADLs or 0.4-km mobility, hearing impairment, visual impairment, grip strength impairment, and cognitive impairment), and psychosocial characteristics (depression36 and social support).
Missing data on covariates, ranging from less than 1% to 16% (mobility status), were addressed using multiple imputation by chained equations.42 Hosmer-Lemeshow tests were applied to multivariable-adjusted logistic regression models to confirm goodness of fit; Pearson goodness-of-fit statistic was used for the negative binomial model.
To account for the competing risk of death during follow-up in the SILVER-AMI study (10% at 6 months after discharge), we conducted secondary analyses of our primary outcomes via the approach recommended by Murphy et al,43 in which we simulated extreme outcomes for decedents (eg, all decedents experienced the outcome or no decedents experienced the outcome), in addition to standard multiple imputation, to evaluate the robustness of our associations of mobility impairment and functional outcomes to bias from death. Analyses were performed using statistical software (Stata 14; StataCorp LLC). Statistical significance was set at 2-tailed P < .05.
Of the 3041 participants in the study sample, 35 died during the index hospitalization, 266 died before the 6-month interview, and 153 were lost to follow-up for reasons other than death (5.6% loss to follow-up), leaving 2587 hospitalized patients with AMI who were 75 years or older with 6-month assessments (Figure 1). Twenty-nine participants were impaired in all 4 ADLs at baseline, and 809 participants could not walk 0.4 km at baseline; these participants were excluded from analyses of ADL decline and decline in 0.4-km mobility, respectively. Excluded participants did not differ significantly in terms of sex or AMI type but were slightly older, less likely to be of white race/ethnicity or to have undergone percutaneous coronary intervention, and more likely to be functionally impaired at baseline.
The mean (SD) age of the sample was 81.4 (4.8) years, and more than half of the sample (56.5% [1462 of 2587]) were male. Eighty-nine percent (2296 of 2587) were of white race/ethnicity, and 43.4% (1114 of 2587) were educated beyond high school. Almost three-quarters of the sample (72.6% [1881 of 2587]) were initially seen with non–ST-segment elevation myocardial infarction, and slightly more than one-third (35.4% [916 of 2587]) were impaired in ADLs or 0.4-km mobility before admission.
Thirty-one percent of the sample (n = 800) were free of mobility impairment (ie, completed the TUG in ≤15 seconds) as measured during the in-hospital assessment (Table). The remainder of participants had mild impairment (564 of 2587 [21.8%]), moderate impairment (414 of 2587 [16.0%]), or severe impairment (391 of 2587 [15.1%]). Differences in characteristics among participants according to mobility status are listed in the Table. Participants with impaired mobility were older and more likely to be female, be of nonwhite race/ethnicity, and live alone (Table). Participants with impaired mobility were also more likely to initially be seen with non–ST-segment elevation myocardial infarction and had higher rates of comorbidities, complications, and impairments in physical, cognitive, and psychosocial characteristics (Table).
One-quarter (648 of 2587) of participants reported functional decline at 6 months after discharge (compared with 4 weeks before admission), 12.8% (332 of 2587) reported ADL decline, and 16.7% (431 of 2587) reported decline in 0.4-km mobility. The most common ADL in which participants declined was bathing (8.5% [219 of 2587]), followed by dressing (6.5% [169 of 2587]), transferring (4.4% [113 of 2587]), and walking around the home (4.1% [105 of 2587]). Among participants who experienced decline, 67.4% (433 of 642) declined in 1 activity, 17.6% (113 of 642) declined in 2 activities, 6.5% (42 of 642) declined in 3 activities, and 8.4% (54 of 642) declined in 4 or 5 activities.
Mobility impairment during hospitalization was associated with decline in ADLs and 0.4-km mobility at 6 months after discharge. Only 3.8% (30 of 800) of participants with preserved mobility reported ADL decline at 6 months after discharge compared with rates of 6.9% (39 of 564) among participants with mild impairment, 18.6% (77 of 414) among participants with moderate impairment, and 34.7% (136 of 391) among participants with severe impairment (P < .001) (Figure 2). Compared with participants with preserved mobility and after adjustment for demographic, clinical, other physical and cognitive impairments, and psychosocial characteristics, adjusted odds ratios (aORs) for any ADL decline were 1.24 (95% CI, 0.74-2.09) for mild impairment, 2.67 (95% CI, 1.67-4.27) for moderate impairment, and 5.45 (95% CI, 3.29-9.01) for severe impairment (Figure 3 and eTable 1A in the Supplement). For decline in ability to walk 0.4 km, aORs were 1.51 (95% CI, 1.04-2.20) among participants with mild impairment, 2.03 (95% CI, 1.37-3.02) among participants with moderate impairment, and 3.25 (95% CI, 2.02-5.23) among participants with severe impairment relative to preserved mobility (Figure 3 and eTable 1B in the Supplement). Sensitivity analyses accounting for the competing risk of death were consistent with the main analyses (eFigure 1 in the Supplement).
We also examined the association of mobility impairment with the count of activities in which each participant declined (range, 0-5). Greater severity of mobility impairment was associated in a dose-response manner with increasing counts of functional decline after AMI, ranging from an incident rate ratio of 1.30 (95% CI, 1.02-1.65) for mild impairment to 2.83 (95% CI, 2.16-3.72) for severe impairment in multivariable-adjusted analyses (eTable 2 in the Supplement).
The incidence of decline in each functional activity according to mobility status is shown in Figure 2. Incidences of decline in each activity were monotonically higher across worsening levels of mobility impairment. For example, only 1.9% (15 of 800) of participants with preserved mobility declined in ability to bathe compared with incidence rates of decline in bathing ability of 4.8% (27 of 564) among participants with mild impairment, 12.8% (53 of 414) among participants with moderate impairment, and 21.9% (86 of 391) among participants with severe impairment.
Multivariable-adjusted ORs for the association of mobility impairment and risk of decline in each essential ADL are shown in eFigure 2 in the Supplement. In adjusted analyses, severe impairment was consistently associated with risk of decline in each activity, with the greatest risk observed for declines in bathing (aOR, 5.28; 95% CI, 2.75-10.14) and dressing (aOR, 4.53; 95% CI, 2.17-9.45). Moderate impairment was associated with significantly increased risks of decline in bathing (aOR, 3.07; 95% CI, 1.61-5.84) and dressing (aOR, 3.24; 95% CI, 1.59-6.60). Mild impairment was not significantly associated with decline in individual ADLs.
In this multicenter prospective cohort study, we found that mobility impairment, measured during AMI hospitalization with the TUG, was associated with increased risk of functional decline at 6 months after discharge. More than half (1369 of 2587) of our older adult cohort exhibited impaired mobility during hospitalization, and impaired mobility was consistently associated with decline in several functions encompassing “essential” ADLs and neighborhood-level mobility. Our findings suggest that the TUG score may be a useful “geriatric biomarker” for identifying older patients with AMI at risk for functional decline.
Mobility status, including gait and balance, is emerging as an important predictor of outcomes among older adults with cardiovascular disease. A recent study20 found slow gait speed to be associated with a nearly 2-fold risk of mortality and hospital readmissions among older patients with AMI. Another study44 reported the TUG score to be a better predictor of complications and mortality after cardiac surgery than established surgery risk scores. The TUG score is considered a surrogate marker for frailty,45 a strong risk factor for clinical outcomes in older patients with AMI.46 The present work adds evidence to the utility of inpatient mobility, and the TUG score specifically, as a marker for poor patient-centered outcomes in older patients with cardiovascular disease.
We found that mobility impairment was associated with decline in all activities we evaluated, including the essential ADLs and neighborhood-level mobility. Mobility impairment showed the greatest risk in its association with declines in bathing and dressing; this finding is important because the loss of ability to independently perform these ADLs may signal a critical transition for older adults from independence to dependence.47-49 Losses of independence in bathing and dressing are associated, more so than loss of independence in other ADLs, with increased risk of hospital admissions50 and institutionalization51; therefore, a brief tool (eg, the TUG) that identifies patients at risk of decline in these domains could be useful in tailoring treatment plans to prevent poor clinical outcomes.
This study is strengthened by the use of data from the largest prospective cohort study to date of patients 75 years or older hospitalized for AMI, recruited from a nationwide network of academic and community hospitals. The SILVER-AMI study rigorously collected a rich array of demographic, cardiac, and geriatric data, allowing us to examine new risk factors in this population while accounting for important traditional risk factors. Follow-up was complete in 94.4% (2587 of 2740) of participants who survived to 6 months, limiting the risk of attrition bias. We also robustly accounted for the competing risk of death in secondary analyses.
These strengths are balanced by some limitations. We operationalized “baseline” functional status as participants’ report of function 1 month before hospitalization. As such, we are unable to precisely identify whether the functional decline reported by participants at 6-month follow-up first occurred immediately before AMI hospitalization, during hospitalization, or afterward. Prior evidence has shown that retrospective report of premorbid functional status (ie, from a time before onset of illness) is a better indicator of baseline function than functional status during hospitalization and is associated with posthospitalization outcomes.52 We acknowledge that some participants may have experienced functional decline around the time of their hospitalization for AMI but subsequently recovered before the 6-month assessment. Other researchers have reported rates of functional decline closer to 30% among older adults shortly after hospitalization, with one-third recovering by 6 months after discharge.53 The 25% rate of functional decline reported in the present study is likely an underestimation of the incidence of functional decline, which potentially biases our findings toward the null and focuses our inquiry on patients with persistent, unrecovered functional decline.
Because most older adults value function and independence as health outcomes of greater priority than longevity,54 clinicians caring for older adults with AMI must recognize and address threats to these important patient-centered outcomes. Our findings suggest that mobility assessment during hospitalization may be useful for evaluating risk of functional decline among older patients with AMI. The TUG is a brief, freely available, and easy assessment that can be administered by clinicians or support staff without extensive training. The assessment requires minimal equipment (a chair with arms, a 3-m [10-ft] floor span, and a clock), all of which are already available in most hospital rooms. We advocate for all clinicians caring for older patients with AMI to administer assessments of balance and gait, such as the TUG.
Despite mobility status demonstrating utility as a potent marker of risk for functional decline, the course of action for mitigating this risk remains elusive. There has been a surge in interest regarding progressive mobilization interventions in the hospital, which show evidence of being safe and feasible for older hospitalized patients, including those with cardiovascular disease.55 However, a recent trial56 of inpatient mobilization found no difference in postdischarge ADLs between experimental and control groups. Other interventions, such as cardiac rehabilitation57,58 and postdischarge case management,59 have shown modest results for improving physical function after AMI, but the low participation rates60-62 of older adults in these programs, particularly those with mobility impairment,63 suggest that further work is needed to develop effective interventions that are suited to the needs and preferences of older adults.
This study’s findings showed that mobility impairment was common among older adults hospitalized for AMI and was associated with risk of functional decline at 6 months after discharge. The TUG is a brief and easy-to-administer point-of-care mobility assessment that may be a useful tool to identify older hospitalized adults at risk of functional decline after AMI.
Accepted for Publication: July 28, 2019.
Corresponding Author: Alexandra M. Hajduk, PhD, MPH, Department of Internal Medicine, Yale School of Medicine, 333 Cedar St, PO Box 208025, New Haven, CT 06520 (firstname.lastname@example.org).
Published Online: October 7, 2019. doi:10.1001/jamainternmed.2019.4114
Author Contributions: Drs Hajduk and Chaudhry had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Drs Hajduk and Murphy and Ms Tsang were responsible for data analysis.
Concept and design: Hajduk, Murphy, Dodson, Gill, Chaudhry.
Acquisition, analysis, or interpretation of data: All authors.
Drafting of the manuscript: Hajduk, Dodson, Haghighat, Chaudhry.
Critical revision of the manuscript for important intellectual content: Hajduk, Murphy, Geda, Dodson, Tsang, Tinetti, Gill, Chaudhry.
Statistical analysis: Hajduk, Murphy, Tsang.
Obtained funding: Hajduk, Murphy, Tsang.
Administrative, technical, or material support: Geda, Dodson, Tsang, Chaudhry.
Conflict of Interest Disclosures: Dr Chaudhry reported receiving grants from the National Institutes of Health and personal fees from CVS. No other disclosures were reported.
Funding/Support: This research was supported by grant R01 HL115295 from the National Heart, Lung, and Blood Institute. This work was conducted at the Yale Claude D. Pepper Older Americans Independence Center (grant P30 AG021342). Dr Hajduk was supported by training grant T32 AG019134 from the National Institute on Aging and by Career Development Award 17MCPRP33670631 from the American Heart Association. Dr Dodson was supported by Patient-Oriented Career Development Award K23 AG052463 from the National Institute on Aging. Dr Gill is the recipient of Academic Leadership Award K07 AG043587 from the National Institute on Aging.
Role of the Funder/Sponsor: The funding sources had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.