The Trojan Lifetime Champions Health Survey: Development, Validity, and Reliability

Participation in competitive sports presents unique and important health considerations. Elite athletes are celebrated for their extraordinary physical achievements and possess superior physiologic characteristics that are associated positively with health (eg, cardiorespiratory fitness, strength, and power) compared with the general population.^1,2 However, competitive sports also are recognized as a potential health risk. Concern for athlete health and safety has led to recent public scrutiny of the long-term consequences of sport participation, notably regarding orthopaedic injury,³ cardiovascular disease,⁴ head injury,⁵ and related psychosocial effects.⁶ Consideration of these risks has been substantial enough to prompt formal inquiries by the US Congress^7,8 and legal action⁹ and to drive extensive changes to health policy.¹⁰ A sense of responsibility is growing among institutions (eg, colleges and universities) and organizations (eg, National Collegiate Athletic Association [NCAA] and professional sports leagues) to understand and promote athletes' health. Despite this attention, data on health and exercise outcomes across the lifespan of elite athletes remain sparse, and a critical need for additional research in this area exists.

Self-report questionnaires are perhaps the most practical way of gathering such data among large, diverse populations. These instruments have been used to describe health and exercise outcomes among select groups of elite athletes.^11–16 However, health, exercise, and health-related quality of life (HRQL) rarely have been comprehensively addressed together. An additional limitation is that most of these instruments have not undergone formal characterization of their psychometric properties in an elite, competitive athletic population. Assessing the validity, test-retest reliability, and precision/error of any instrument in relation to clinically meaningful effects is critical to establishing confidence in study results and their implications for health care and policy.¹⁷

As part of the collaborative research and education program Trojan Lifetime Champions (TLC), our research team developed a comprehensive instrument to measure lifetime and recent health, exercise, and HRQL among current and former university students, including collegiate athletes. Our ultimate aim is to use the questionnaire to better understand the unique influence of elite competitive sports on lifespan health and well-being. Therefore, the purpose of our study was to evaluate the validity and reliability of a novel health and exercise questionnaire: the TLC Health Survey. We describe the development and formal psychometric assessment of that instrument, with the goal of providing an accumulation of evidence to support its use.¹⁷

METHODS

The TLC Health Survey

Overview

The TLC Health Survey includes 216 unique items in 5 sections (Demographics, Experience in Competitive Sports, Health Assessment, Health-Related Quality of Life, and Current Exercise Behavior / Health & Exercise Attitudes) as presented in Table 1. It incorporates both an established instrument and novel measures of interest. The survey initially was a paper questionnaire, with a Web-based electronic version developed in parallel. The electronic survey contained identical content presented in a series of Web pages that matched the pages of the paper survey and was administered via a dedicated Web site. The TLC Health Survey is available online (see Supplemental Material found at http://dx.doi.org/10.4085/1062-6050-50.2.10.S1).

Table 1. Trojan Lifetime Champions Health Survey Content

View Fullscreen

Development Process

We sought an instrument to provide comprehensive data on lifetime and recent holistic health, exercise behavior, and HRQL among a diverse population of current and former University of Southern California (USC) students, including elite, competitive athletes. After conducting a literature review and a series of informal, open-ended interviews with former athletes and consulting with knowledge experts, we determined that no existing validated instrument provided a suitable means to do so. In particular, health assessments were focused too narrowly on specific regions (eg, knee) or conditions (eg, osteoarthritis), and exercise instruments did not capture important features of exercise behavior (eg, resistance training). Therefore, we developed a novel questionnaire for our study.

Development of the TLC Health Survey was a collaborative process involving an interdisciplinary team of health and exercise scientists and diverse experts affiliated with the USC Department of Athletics (Table 2). Through a series of group meetings, 6 major survey revisions, and 4 pilot administrations in different populations, we built an instrument that the study team deemed to possess appropriate content validity and suitability for large-scale administration. This development process is detailed in Table 3.

Table 2. Trojan Lifetime Champions Health Survey Development Team

View Fullscreen

Table 3. Trojan Lifetime Champions Health Survey Development Timeline, Major Revisions, and Pilot Studies

View Fullscreen

At each stage of the development process, we revised the survey to include comprehensive content, improve clarity and usability for the survey taker, and minimize opportunities for bias. We incorporated color-coded instructions and examples, along with careful language and the specific criteria used to evaluate the prevalence and severity of various health concerns. A standardized set of guidelines was developed for interpretating paper-survey responses. These guidelines provided specific rules that ensured consistency of data entry when responses were ambiguous, contradictory, or missing. For example, if age was reported as a decimal or fraction (eg, 19.5), it was rounded down to the nearest integer (eg, 19). Interrater reliability was assessed by repeated entry of a random 10% sample of surveys, with excellent (99.96%) repeatability. The complete set of guidelines is available online (see additional Supplemental Material found at http://dx.doi.org/10.4085/1062-6050-50.2.10.S2).

Feedback from the interdisciplinary study team, and particularly the expertise of a licensed clinical psychologist (R.M.S.), was essential to developing a user-friendly instrument. The collective expertise of this group, their knowledge and experience with an elite athletic population, and systematic documentation of the development process provided substantial evidence for content validity.^17,18

Section I: Demographics

Demographic information, consisting of age, sex, ethnicity, height, and mass, was recorded in a series of straightforward multiple-choice and short-response questions typical to epidemiologic instruments.

Section II: Experience in Competitive Sports

History of participation in Division I collegiate athletics, consisting of ages at times of participation, sport or sports played, and in-season status, was recorded in a series of yes/no, multiple choice, and short-response questions. In addition, self-rated perception of the influence of competitive sports on overall health and the record of any postcollegiate professional sports participation were documented.

Section III: Health Assessment

A lifetime and recent health inventory was developed uniquely for this instrument. The Section III health inventory assessed the prevalence of health concerns across 6 domains: Joints, Bone & Muscle, Cardiopulmonary, Neurological, Other Clinical, and Psychosocial. Other Clinical included items that did not clearly fall into other domains (eg, liver, kidney, and lymphatic system). A total of 59 unique health items were identified (Joints = 11, Bone & Muscle = 9, Cardiopulmonary = 8, Neurological = 9, Other Clinical = 9, and Psychosocial = 13). The health inventory was structured as a set of matrixed tables, with a table for each health domain. The Joints domain table is provided as an example in Figure 1. For each item in the table, respondents described lifetime and recent (3 years before the study) health concerns. We quantified concerns using a 4-point ordinal scale calibrated to the degree of professional treatment, as well as the age at which individuals first experienced symptoms. Choices were No Concerns, Some Concerns (without attention from a professional), Serious Concerns (treated by a medical professional), and Major Concerns (with surgery or hospital stay). Concerns were defined with respect to a specific body part (eg, knee) and without discrimination of specific pathologic conditions (eg, ligament versus cartilage injury). We adopted this structure to establish straightforward, patient-centered quantitative criteria for describing health concerns, thereby simplifying the analysis while reducing the opportunities for response bias and outcome misclassification.

View Figure

• Download as Powerpoint
• Download Figure

Summary scores for each domain of the health inventory (domain scores) were calculated by summing individual item scores from the domain and were computed for both lifetime and recent experience. Lifetime domain scores reflect cumulative lifetime experience with health concerns in a given domain, and recent domain scores reflect cumulative experience for the 3 years before the survey. A summary domain score of 0 indicates perfect health in that domain, and progressively higher scores indicate greater evidence of concerns. Maximal possible scores ranged from 24 for the 8-item Cardiopulmonary domain to 39 for the 13-item Psychosocial domain.

Section IV: Health-Related Quality of Life

The Short-Form 12 Version 2 Health Survey (SF-12v2) was adapted¹⁹ for Section IV (HRQL) and is referred to as SF-12 throughout this document. The TLC Health Survey used identical text, but we reformatted fonts and layout to match the remainder of the questionnaire. The SF-12 and its parent instrument (the Short Form-36) have been widely used to measure HRQL across diverse populations, including collegiate^15,20–22 and former professional athletes.²³ The inclusion of an established, validated, and recognized instrument, such as the SF-12, enhances content validity.^17,18 The SF-12 Physical Component scores (PCS) and Mental Component scores (MCS) were computed from a proprietary algorithm for summary physical and mental health.¹⁹ A score of 50 reflects the approximate population mean for all US adults, with higher scores indicating better HRQL.

Section V: Current Exercise Behavior / Health & Exercise Attitudes

Exercise behavior questions were designed to mirror weekly exercise guidelines for healthy adults as published by the American College of Sports Medicine (ACSM).²⁴ We were unaware of a previously validated instrument specifically designed to measure acute exercise behavior, including resistance exercise, relative to these guidelines. Specifically, the TLC Health Survey measures the previous week's frequency and average duration of exercise across 3 activity subtypes: cardiovascular, resistance, and mixed. Cardiovascular exercise included activities such as brisk walking, hiking, running, swimming, using an elliptical, and cycling. Resistance exercise included weight lifting with or without machines. Mixed exercise encompassed all other exercise activities, including sports (eg, basketball, soccer, volleyball) and heavy yard work. A minimal exercise threshold of moderate intensity, which was defined as “working hard enough to raise your heart rate and break a sweat, yet still being able to carry on a conversation,” was specified for all forms of exercise. Exercise volume (in minutes) from the previous week was calculated for each form of exercise as the reported number of sessions multiplied by the average session duration. Total exercise volume from the previous week was calculated as the sum of exercise volumes across the 3 exercise subtypes. Given our definition of the minimal exercise threshold of moderate intensity, we reasoned that all reported exercise (cardiovascular, resistance, and mixed) contributes to target cardiorespiratory exercise training volume as specified by the ACSM guidelines.²⁴ Self-reported exercise importance was measured using a multiple-choice item with a 4-point ordinal scale. The question read, “In general, how important would you say exercise is in your life?” Response choices were Very Important, Important, Somewhat Important, and Not Important. Several multiple-choice questions regarding health-and-wellness perceptions and exercise attitudes also were incorporated into Section V.

Study Population and Recruitment

We divided USC alumni into 2 subgroups: athlete alumni and nonathlete alumni. To be eligible for participation, athlete alumni must have practiced or competed in Division I athletics at USC. Nonathletes were former undergraduate students at USC who never practiced or competed in Division I athletics at any university.

Prospective study participants were recruited through university records, student and alumni organizations, and personal referrals. Study participants were selected to provide a convenience sample representative of the USC athlete alumni source population and an age- and sex-matched control group. For each current varsity sport, we examined media guides to estimate roster sizes over the previous 50 years and defined accordingly the demographic characteristics of the source population, including sex, age, and sport. An estimated 4500 USC varsity athlete alumni were identified. Characteristics of the source population are presented in Tables 4 and 5. Athlete alumni were targeted to proportionally match these demographic characteristics.

Table 4. Demographic Characteristics of Participants in the Trojan Lifetime Champions Phase 2 Validity and Reliability Study, 2010–2011, Compared With the University of Southern California Varsity Athlete Alumni Source Population

View Fullscreen

Table 5. Athlete Participants' Sports in the Trojan Lifetime Champions Phase 2 Validity and Reliability Study, 2010–2011, Compared With the University of Southern California Varsity Athlete Alumni Source Population

View Fullscreen

Nonathlete alumni were recruited to mirror the athlete alumni based on age (primarily) and sex (secondarily). Nonathletes were excluded from participation if they indicated any experience in varsity or organized club sports during college. Those with experience in high school or collegiate intramural sports were allowed to participate. Control participants who were matched only on age (but not sex) were also included.

Study recruitment and participation are shown in Table 6. A total of 109 participants (58 athletes and 51 nonathletes) were recruited for the study (recruit population). Of these, 86 (79%) completed the first survey (study population), and 63 (58%) completed both surveys (reliability population). The 63 participants in the reliability population were included in the subsequent survey validity and reliability analysis. All participants provided informed consent, and the experimental protocol was approved by the USC Health Sciences Institutional Review Board.

Table 6. Trojan Lifetime Champions Phase 2 Validity and Reliability Study, 2010–2011, Recruitment and Participation

View Fullscreen

RESULTS

Response rates were similar between athlete and nonathlete alumni. Responders and nonresponders were similar in sex (χ²₂ = 0.91, P = .64), age (χ²₁₀ = 10.0, P = .44), collegiate sport participation (χ²₂ = 0.83, P = .66), and survey type (χ²₆ = 3.51, P = .74). Age (t₃₂ = −0.259, P = .80) and sex (P = .40) distributions were similar between the athlete source and reliability populations (Table 4). We found a difference in sport distribution between the athlete source and reliability populations (χ²₁₈ = 43.6, P < .001; Table 5). Specifically, the reliability population overrepresented athletes in women's basketball, men's golf, women's rowing, men's swimming and diving, and men's tennis and underrepresented athletes in women's cross-country, men's football, men's and women's track and field, and women's water polo. Two multisport athletes were part of the reliability population, including a man who played both basketball and football and a woman who played both golf and tennis.

The mean test-retest interval was 23.1 ± 30.2 days. This was longer than the self-directed interval of 1 week (t₆₂ = 4.22, P < .001). We found no effect of collegiate sport participation (t₆₁ = 1.071, P = .29) or survey type (F_3,59 = 0.41, P = .75) on test-retest interval. Women (14.5 ± 13.0 days) had a shorter test-retest interval than men (27.1 ± 34.9 days; t₅₉ = −2.07, P = .04), with unequal variances per the Levene test (F_1,61 = 7.67, P = .007).

Subsequent to outlier screening, data from 4 participants were excluded from analysis in the exercise-behavior variables. Two participants were excluded from analysis for each of the Bone & Muscle and Neurological domains. Figure 2 illustrates a sample of outlier data.

View Figure

• Download as Powerpoint
• Download Figure

Results of the test-retest and parallel-form reliability assessments are summarized in Table 7. Test-retest reliability coefficients (ICC) averaged 0.86, 0.90, 0.80, and 0.74 for HRQL, lifetime domain scores, recent domain scores, and exercise variables, respectively, indicating very large to extremely large agreement.²⁵ The Bone & Muscle recent domain score and mixed exercise variables were notable exceptions, with ICCs of 0.58 and 0.49, respectively. These values reflect moderate to large agreement.²⁵ The Bone & Muscle recent domain score showed a substantial difference between ICCs for PP (0.78) and EE (0.50) survey types.

Table 7. Test-Retest and Parallel-Form Reliability Results for Participants in the Trojan Lifetime Champions Health Survey Phase 2 Validity and Reliability Study, 2010–2011

View Fullscreen

Parallel-form reliability coefficients (ICC) averaged 0.90, 0.85, 0.85, and 0.81 for HRQL, lifetime domain scores, recent domain scores, and exercise variables, respectively, indicating very large to extremely large agreement. Substantial differences between ICCs for PE and EP survey types were noted for the SF-12 PCS score (0.65 versus 0.92), Joints lifetime domain score (0.67 versus 0.90), and Psychosocial recent domain score (0.70 versus 0.92). Each of these correlation coefficients, however, demonstrated large (or greater) agreement. For the Bone & Muscle lifetime domain score, the PE survey type had an ICC of 0.27, which was not statistically significant, whereas the EP survey type had an ICC of 0.91, which was statistically significant.

Systematic error was detected for the Other Clinical lifetime domain score, with an average test-retest difference of −0.48 ± 1.39 units. No other variable demonstrated evidence of a systematic change.

Typical errors were 2.85, 1.12, and 0.93 units for HRQL, lifetime domain scores, and recent domain scores, respectively. For exercise volume of the previous week, typical errors were 58.7, 17.2, 74.5, and 79.7 minutes for cardiovascular, resistance, mixed, and total exercise, respectively. Typical error for self-rated exercise importance, modeled as a continuous variable, was 0.30 units. In all cases, 95% confidence limits for typical error were smaller than respective minimal a priori thresholds for substantial clinical meaning.

Results of the convergent construct validity assessment are summarized in Table 8. Negative Pearson product moment correlations were observed between 5 of the 6 lifetime and recent domain scores, excluding the Psychosocial domain, and the SF-12 PCS HRQL score. Coefficients ranged from −0.33 to −0.65 for lifetime domain scores and from −0.49 to −0.69 for recent domain scores, suggesting moderate to large associations.²⁵ The Psychosocial lifetime (r = −0.58) and recent domain (r = −0.62) scores were correlated with the SF-12 MCS HRQL score, reflecting a large association.

Table 8. Convergent Construct Validity Correlation Coefficients for Participants in the Trojan Lifetime Champions Health Survey Phase 2 Validity and Reliability Study, 2010–2011^a

View Fullscreen

DISCUSSION

We conducted a systematic validity and reliability assessment of a novel health-and-exercise questionnaire (the TLC Health Survey) designed to measure lifetime and recent health, exercise, and HRQL among elite, competitive athletes. The study was conducted in a population of former varsity collegiate athletes and an age- and sex-matched control group of alumni at a large NCAA Division I university. The TLC Health Survey demonstrated excellent test-retest reliability and parallel-form reliability (ie, equivalence) between the paper and electronic versions across a diverse range of measures, including 6 domains of lifetime and recent health, several forms of exercise, and HRQL. Bone & Muscle domain scores were less reliable than other data and should be interpreted with caution. Quantitative estimates of measurement error were less than a priori thresholds of clinical interest. The instrument demonstrated strong evidence of convergent validity in relation to the SF-12. The development and implementation process of the instrument, as well as inclusion of an established HRQL instrument, reflect evidence for content validity and feasibility of administration in the populations of interest.

These findings provide strong evidence of test-retest reliability and parallel-form reliability between the paper and electronic versions of the TLC Health Survey. In general, ICCs for TLC domain scores exceeded 0.70. Lifetime domain score ICCs generally exceeded 0.80, whereas ICCs for recent domain scores were somewhat lower. Reliability and validity for these novel measures were comparable with those of the widely used SF-12 PCS and MCS HRQL indices.

Lower correlation coefficients and evidence of substantial differences between survey types were found within the Bone & Muscle domain; thus, these scores should be interpreted with caution. Instructions for this domain possibly lacked clarity when compared with instructions for the other domains. The Bone & Muscle instructions directed participants to “describe bone & muscle concerns [including] fractures, pulled muscles or tendons, pain, stiffness, or weakness” in various regions of the body (eg, lower leg, upper torso, and hand). In retrospect, these instructions may be less specific and, therefore, more subject to misclassification than the reference to a particular joint (eg, knee) or medical condition (eg, high blood pressure) described in other domains. In future studies, researchers might consider identifying specific bones and muscles (eg, clavicle, hamstrings) or combining joints, bones, and muscles into a single musculoskeletal domain within the instrument. Bone and muscle concerns also could be more transient or episodic,²⁸ leading to differential reporting between the test and retest that appears as measurement error.

We found some evidence of an influence of test order in the parallel-form reliability assessment. The ICCs differed substantially (>0.2) between the PE and EP survey types for the SF-12 PCS, Joints lifetime domain score, and Psychosocial recent domain score. However, the overall interpretation of large (or greater) agreement between the test and retest was unaffected, and we do not believe this presents a substantial threat to reliability for these measures. Authors of future studies in expanded sample populations might explore whether this effect has a systematic explanation. Clinicians and researchers using multiple iterations of the TLC Health Survey should consider using the same form (paper or electronic) for all administrations.

Validity and reliability of exercise-outcome variables generally were excellent. Mixed Exercise, which by its nature is perhaps more ambiguous than Cardiovascular or Resistance Exercise, had relatively lower ICCs but still demonstrated moderate agreement. Exercise behavior is more likely to vary from week to week than measures of health over the lifespan or the previous 3 years (ie, recent domain scores). Thus, changes in exercise outcomes between the test and retest that appear as measurement error actually may be attributable to true changes in behavior. Exercise data constituted the most common exclusion in the outlier screening process. Figure 2 depicts 2 examples of outliers for weekly Cardiovascular Exercise. One participant reported nearly 19 hours at both the test and retest and approximately 50 total hours of exercise per week. These values are questionable and well outside the group data. Another participant reported more than 13 hours in the first survey and just 8 minutes in the second. This likely reflects week-to-week variability as opposed to measurement error. In the practical effort to measure exercise using self-report questionnaires, these challenges remain an important limitation for this and any comparable instrument.

Whereas no universally accepted consensus for acceptable validity and reliability measures exists,^25,29 the test-retest and parallel-form reliability coefficients that we report are comparable with those in previous studies of the SF-12^30–33 and SF-36³⁴ HRQL instruments. Similarly, reliability for the exercise measures was comparable with that of the highest-quality physical activity instruments identified in a recent systematic review.³⁵ Nonetheless, given that this is one of the first formal assessments of validity and reliability for a health-and-exercise outcomes instrument in an elite, competitive athletic population, additional larger-scale studies in similar populations are warranted to provide further validity evidence.

Review of the test-retest differences demonstrated minimal evidence of systematic error. One exception (Other Clinical lifetime domain score) demonstrated a reduction, but the magnitude of this change was less than one-fourth of the minimal a priori threshold for substantial clinical meaning. In all cases, typical error was less than the respective threshold of interest. More powerfully than a correlation coefficient, these data provide evidence that observed differences in outcome variables between groups or over time reflect true differences of clinical interest as opposed to measurement error. This “magnitude-based” approach is a preferred alternative to traditional hypothesis testing versus the null value.²⁵

We find it interesting that the mean test-retest interval was more than 3 times the prescribed 1-week interval and was different between men and women. Although this illustrates a limitation to study compliance, an interval of 3 weeks is not uncommon for reliability studies of comparable instruments. The SF-12, for example, has undergone test-retest reliability assessment at intervals ranging from 1 week to 3 months with comparable results.^30–32 Whereas women had a shorter test-retest interval than men in this study (approximately 2 versus 4 weeks), reliability coefficients were similar. A longer test-retest interval increases the potential for true changes in health status or exercise behavior that appear as measurement error. One-week reliability, therefore, might be presumed superior to that measured at 3 weeks. However, a reliability study for a musculoskeletal-symptoms questionnaire showed no differences between retest intervals of 2 and 4 weeks.³⁶

The correlation of questionnaire items sharing the same theoretic construct provides evidence of construct validity.¹⁷ Our findings indicate moderate to large associations among each of the TLC Health Survey physical health domain scores and the SF-12 PCS physical HRQL index, as well as the Psychosocial health domain scores and the MCS mental HRQL index. Correlations were somewhat larger for recent domain scores than for lifetime domain scores. This likely is attributable to the fact that SF-12 scores also reflect recent (past 4 weeks) health experience. Although health domain scores (measures of health outcomes) and SF-12 HRQL scores share broad constructs of physical and mental health, they do not measure exactly the same components of health. Thus, a strong correlation would not be expected. However, the reported associations demonstrate both validity evidence for the novel health domain scores and the relevance of association with HRQL.

The primary limitation of this study was the relatively small sample population compared with source populations of interest. More than 450 000 individuals actively participate in NCAA sports, a figure that has increased more than 60% over the past 20 years.³⁷ Whereas this study population was representative of the age and sex distribution of former athletes at 1 Division I university, it does not necessarily represent the overall athlete population or that of other elite competitive athletes (eg, professionals and Olympians). Furthermore, we found a difference in sport distribution between the source and study populations, but the sample size was insufficient to evaluate sport-specific influences on the findings. Together, these factors limit the generalizability of study results and support the need for additional studies with larger and more diverse sample populations. Additionally, this instrument was developed to assess health and exercise in athletic populations, but it could be readily adapted to other populations (eg, soldiers, firefighters) in whom demands for optimal physical performance present comparable lifespan health challenges. Independent validity assessments would be required to confirm the suitability of the instrument in these populations. In general, these findings should be considered a substantial first step in the ongoing accumulation of evidence necessary to support the utility, validity, reliability, and generalizability of this novel instrument.¹⁷

Evaluation of psychometric properties using the robust measures presented here distinguishes the TLC Health Survey from previous questionnaires used to assess health outcomes and exercise behaviors in comparable populations and, indeed, from most epidemiologic instruments. Bennett et al³⁸ reported that of 117 recently published reports of self-administered surveys, less than 20% provided validity or reliability data for the instrument. Similarly, less than 30% of physical activity instruments reviewed by van Poppel et al³⁵ met the highest level of reliability evidence. Whereas such evidence typically is given in the form of test-retest correlation coefficients, researchers^25,29 have argued that the typical error and change in the mean between trials are more important measures. Our study provides both correlation coefficients and estimates of error magnitude relative to thresholds of substantive meaning.²⁵ It provides evidence of content validity as determined by an interdisciplinary team of experts and through inclusion of an established instrument in the SF-12. Furthermore, it provides assessments of feasibility and external validity in relation to source, study, and target populations of interest. Finally, parallel-form reliability between the paper and electronic versions supports the use of the electronic survey as an equivalent instrument in large-scale studies, offering an additional feasibility benefit.

CONCLUSIONS

Our study provides strong evidence for the validity and reliability of the TLC Health Survey in assessing lifetime and recent health, exercise, and HRQL among university alumni, including NCAA Division I athletes. The formal characterization of its psychometric properties, including the evaluation of error magnitude relative to thresholds of substantial clinical meaning, distinguishes the TLC Health Survey from previous questionnaires used to evaluate lifetime health and exercise in elite athletes. Additional, larger-scale studies in diverse populations are necessary to enhance the generalizability of the instrument. The electronic form of the TLC Health Survey is a suitable and feasible means to do so.