Abstract
BACKGROUND:
Many epidemiologic studies of vestibular disorders are based on responses to questionnaires which have not been tested against objective tests of the vestibular system.
OBJECTIVE:
The goal was to determine if the dizziness and balance questions used in the National Health Interview Survey (NHIS) are valid and predict performance on objective tests of the vestibular system (VNG).
METHODS:
Data from 367 participants recruited from the community, aged 21.4 to 87.6 years, were collected in the Otolaryngology department at a tertiary care center. They were asked the eight NHIS questions twice, at least 30 minutes apart and were tested on VNG.
RESULTS:
Question responses changed from Test 1 to Test 2 and differed between males and females. “Yes” responses did not predict abnormal VNG responses, for the total group and when the group was categorized into younger (<60 years) and older (>60 years) subjects. The sensitivity and specificity of all questions was low.
CONCLUSION:
The NHIS questions provide some information about what people recall of their experiences, but they may not provide insight into the diagnostic prevalence of vestibular and balance disorders because the sensitivity and specificity are too low. Questionnaire-based epidemiologic studies should be interpreted with caution.
Introduction
The literature on vertigo and balance problems is largely based on questionnaires and survey instruments, including country-specific studies [4, 21] and one multinational study [18]. Only a few studies used questionnaires that were directly compared to diagnostic tests or clinical examination. Von Brevern et al. [19] compared a subgroup of their respondents’ answers to the outcome of clinical examinations to determine how well the questions predicted actual assessments in the clinic [15, 19]. Bisdorff et al. [4] used a scale that had been developed previously [22] but the diagnostic classifications with the previous subjects were unknown. Havia et al. also examined a subset of their sample with clinical examinations [10] but the sample was small. In other words, none of these scales were well-validated.
In the United States, the Centers for Disease Control and Prevention has run the National Health Interview Survey (NHIS) to learn about the health of the American people each year since 1957. In 2008, 2014 and 2016 eight questions about dizziness and balance were included. (See Balance and Dizziness supplement https://www.cdc.gov/nchs/nhis/quest_doc.htm.) Subsequent analyses of the data collected with those questions indicated relationships between balance/ dizziness and co-morbid factors of sex, age, cognition and psychiatric conditions [3] and bilateral vestibular impairment [20]. In particular, history of depression, anxiety disorder and panic disorder were all related to “vestibular vertigo” as indicated by responses on the NHIS questions about true self-motion vertigo, positional vertigo and/ or oscillopsia. Furthermore, respondents tend to conflate descriptions of different types of dizziness [12].
Although asking these questionnaires is very important in population-based epidemiologic research, to date, to the best of our knowledge, no studies have compared the responses on these the questions to the results of objective tests of the vestibular system. We are unaware of any studies performed by the NHIS staff to validate their questionnaire prior to implementation. This lapse presents a significant methodologic problem. It might have been corrected with use of questionnaires used in clinical practice, but for the problem that history-taking in patients who present themselves with “dizziness” which may or may not indicate vestibular disorders is unreliable [17]. Some of the NIHS questions are similar to questions used by neurotologists during the history portion of the initial examination. Recent evidence shows that most questions asked by neurotologists during the history portion of an initial visit are not related to eventual findings on objective vestibular testing [6].
The goal of the present study was to determine the relationship between responses to the NHIS questionnaire and responses on objective vestibular function tests. The NHIS gave the questions only once. We gave the questions twice, at two different time points in the same visit, to learn if responses change when people have more time to think about the questions. In order to simulate the NHIS experience on a small scale, we did not seek out patients with known vestibular disorders from a clinical practice but recruited self-described healthy people from the community.
Methods
Subjects
Subjects in this study were volunteers with no known disorders of the vestibular system, recruited from among staff, students, and healthy visitors, i.e. patients’ companions, and community members such as senior exercise groups. They were all independently ambulatory without canes or walkers. No subjects had disorders of ocular motility. No subjects took medication for anxiety, nausea or other medications with vestibular-suppressant effects. The cohort was comprised of 367 subjects including 262 females, mean age 59.8 yrs., range 21.9 to 87.6 yrs., and 105 males, mean age 48.3 yrs., range 21.4 to 84.5 yrs.
All subjects gave written informed consent prior to participation. This study was approved by the Institutional Review Board for Human Subjects Research at the corresponding author’s institution. Data were collected between February 2017 and October 2019.
Methodology
After giving informed consent, subjects were asked the eight NHIS questions by a staff member. They were allowed as much time as they needed to respond. They were asked the questions again 30 to 90 minutes later, and again were given as much time as they needed to respond. In between the two question sessions they were tested on walking balance with standard tandem walking and on standing balance with Romberg tests and then were tested for 90 minutes on the standard clinical battery of objective diagnostic tests of the vestibular system (VNG), described below. Therefore, subjects were occupied with other testing in between question sessions.
On both the first and second trials of questions the staff member recorded the responses, which were either Yes or No. No other responses were allowed. The questions were as follows: Have you ever had a problem with dizziness, lightheadedness, feeling as if you are going to pass out, unsteadiness or imbalance? Do not include times when you have been drinking alcohol or using other drugs or medications. The following are symptoms of dizziness and balance problems. For each of the following indicate whether or not you have had a problem with any of these symptoms. Do not include times when you have been drinking alcohol or using other drugs or medications. A spinning or vertigo sensation, or other illusion of motion such as tipping, tilting or rocking A floaty, spacey, or disconnected sensation Feeling lightheaded without a sense of motion Feeling as if you are going to pass out or faint Blurring of your vision when you move your head Feeling off balance or unsteady Other dizziness or balance problem
All subjects were given the standard VNG available, including cervical vestibular-evoked myogenic potentials (cVEMP) recorded with silver-silver chloride electrodes, low frequency sinusoidal tests of the vestibulo-ocular reflex in darkness (VOR) in a rotatory chair (Neurokinetics), Dix-Hallpike maneuvers, supine roll tests, and bi-thermal caloric tests using water. During rotatory chair tests, Dix-Hallpike, supine-roll and caloric tests eye movements were recorded with infra-red video-oculography.
On bi-thermal caloric tests the cut-point for abnormal was ≥20% unilateral weakness. For Dix-Hallpike maneuvers and supine roll tests the presence of 3 or more beats of nystagmus was considered abnormal. For cVEMP responses were considered abnormal if responses were absent, or thresholds were 65db. For VOR tests in the rotatory chair we used the norms for gain and phase that were developed by the manufacturer. If a subject had one or more positive responses, i.e. abnormalities, that person was coded as abnormal in the database.
Statistical analyses
To identify which subjects had vestibular system abnormalities we tried to approximate the conditions available at a range of clinical laboratories, which have a range of tests available. Therefore, we examined data from the total VNG battery. Also, because some clinical labs do not perform cVEMP or rotatory chair tests, we looked at the effect of removing those tests, too. The reduced VNG refers to tests without the rotatory chair. The minimal VNG refers to tests without the rotatory chair or cVEMP, i.e. it includes only Dix-Hallpike testing, supine roll testing, and bi-thermal caloric testing.
Subjects’ responses on Trials 1 and 2 were compared by McNemar tests. To provide question-specific sensitivity/specificity, responses on each question were compared with results on the VNG by Chi-square test. Furthermore, responses by males and females on each question were compared by Chi-square test. P < 0.05 was considered statistically significant. Statistical analyses were performed using SAS statistical software (version 9.4; Cary, NC, USA).
Results
When responses to questions on Trials 1 and 2 were compared, McNemar’s tests showed that responses on some, but not all, questions differed significantly. The percentage of Yes responses differed significantly from Trial 1 to Trial 2 on the following questions: Q1 all symptoms, Q2 vertigo, Q2c lightheadedness, Q2d feeling of being about to pass out, Q2f unsteadiness. Therefore, the responses on Trials 1 and 2 were examined separately. See Table 1.
McNemar tests comparing percent of Yes responses on Trials 1 and 2 on each question
McNemar tests comparing percent of Yes responses on Trials 1 and 2 on each question
Using Trial 1, only a few questions had significant Chi square results. We found few significant relationships between the percentage of Yes responses on questions and the percentage of positive responses on the VNG. None of those questions had high sensitivity, in either the total group or the groups broken into younger and older subjects by age 60. See Table 2. Similarly, on Trial 2, we found few significant relationships between the percentage of Yes responses on questions and the percentage of positive responses on the VNG. None of those questions had high sensitivity, in either the total group or the groups broken into younger and older subjects by age 60 for convenience. See Table 3.
Sensitivity, specificity, true and false positives and negatives and p-values for Trial 1 responses with reference to the Total VNG, the Reduced VNG and the Minimal VNG. Scores in plain font are values for subjects < age 60. Scores in italics are values for subjects ≥age 60. Scores in
Sensitivity, specificity, true and false positives and negatives and p-values for Trial 2 responses with reference to the Total VNG, the Reduced VNG and the Minimal VNG. Scores in plain font are values for subjects < age 60. Scores in italics are values for subjects ≥age 60. Scores in
Having the VNG outcomes may be useful for understanding these findings. On sinusoidal tests of the vestibulo-ocular reflex in darkness with all frequencies combined, 9.8% of subjects had abnormal gains and 14.4% of subjects had abnormal phases. On bi-thermal caloric testing, 11.3% of subjects had unilateral weakness. On Dix-Hallpike tests 22.4% of subjects had positive responses. On cVEMP 1.6% of subjects had abnormal responses.
When males and females were analyzed separately the responses on some but not all questions differed significantly. Three questions showed significant differences: 1, all symptoms combined, 2c, lightheadedness, and 2d, feeling faint. See Table 4. No questions had high sensitivity and specificity in males and females separately. See Table 5.
Differences between males and females on Trial 1. Percent of subjects who answered Yes on individual questions. (Females, n = 262, males, n = 105)
Sensitivity and specificity by sex with reference to the Total VNG
The NHIS vertigo and balance questions were never intended by the test developers to diagnose specific vestibular pathologies. To the best of our knowledge this study is the first to compare the outcome of widely-used epidemiologic study questions to the results of objective vestibular function tests. Based on the sensitivity and specificity findings for each separate question, these questions are better at screening for normal responses than abnormal responses. These findings suggest that the responses on the NHIS vertigo and balance questions do not predict results on objective vestibular function tests in the general adult population. This finding is true across age groups, as younger and older subjects had similar responses. Previous work with a quality of life questionnaire for dizziness had a similar finding [23].
Using subjective questions about hearing and objective measures of hearing impairment the United States National Health and Nutrition Examination Survey (NHANES), an on-going, multi-year study of the health of thousands of community volunteers, showed that self-reported hearing impairment is underestimated in older subjects, especially women compared to men, and less accurate in subjects with less education [11]. Choi et al. showed that self-reported hearing impairment was associated with subjective question responses but not with objective audiometric test findings [5].
For these reasons, self-report questionnaires that have not been validated against objective tests of sensory function should not be used as indicators of sensory impairment. In particular, self-reported symptoms that might be consistent with vestibular disorders are not good indicators of actual test data and should not be used as predictors or as a substitute.
For unknown reasons, males and females responded somewhat differently on some of the NHIS questions: Question 1, about all symptoms combined, and Questions 2c and 2d, about lightheadedness and feeling faint, respectively. The differences may have been due to the difference in sample sizes. Alternatively, men and women may have different experiences of lightheadedness and syncope. This difference between men and women is similar to the report by Kamil et al. about hearing impairment [11]. Of note, however, is the finding in a recent study of blood pressure and results of objective vestibular tests that more males than females had high blood pressure [6]. Also, some studies show that women are more likely than men to have low blood pressure [1, 9]. A more detailed exploration of this topic is beyond the scope of this study.
Several ideas explain differences between Trials 1 and 2. Some people may have needed more time to consider their responses and the interval between the two trials may have allowed them to recall how they felt in the past. Some people may have been responding to sensations elicited during testing, either to over- or under-interpret sensations of vertigo. Other people may have had a single episode of vertigo in the past, or episodic attacks of vertigo and/ or other symptoms and their responses may have varied depending on how they interpreted the questions in reference to their own experiences. Whatever the reasons, the responses to these questions are interesting but may not be good indicators of vestibular function in the general population. People may not be reliable enough, as indicated by the relatively higher weight on Question 2g, Other symptoms. Future population-based studies should use more probing questions (e.g. current or past time, chronic vs episodic symptoms) to address this bias, compared to objective test findings.
Community-dwelling people who self-identify as having health problems are more likely to show stronger relationships between the answers to questions about the self-perceived health problem and responses on objective measures than other people in the same community who have not self-identified as having that health problem [2]. This behavior may be a manifestation of the so-called “healthy worker effect” or “community vs. hospital” effect [13, 16]. The differences might involve personality differences, life stressors, or the occurrence of other acute or chronic health conditions. In other words, people in the community may become patients when they have fewer resources with which to maintain their ability to function. Regardless of the reasons, epidemiologists and health care providers should be aware of the problems with these questions. Health professionals who care for patients complaining of vertigo and balance problems often ask questions similar to the questions used in the National Health Interview Survey. We do not know if self-identified patients are more likely to show stronger relationships between the answers to these question and responses on objective diagnostic testing.
The study has some other limitations. We had fewer men than women. Ideally, more men would have been included but we used a convenience sample. The difficulty in recruiting men seems to be a sociological phenomenon of availability and interest. Many middle-aged and older participants were women who saw signs advertising the study when they visited the facility to accompany family members to clinic visits. Recruiting subjects aged 80 or older was challenging. We were unable to recruit more subjects in the time available.
We used a small, convenient sample and tested subjects in a clinical setting. Therefore, extrapolation of these results to large scale epidemiologic studies should be done with caution. As we did in our study, these questions, including probing questions, should be validated with larger samples before they are used in population-based epidemiologic studies.
Funding
This study was funded by NIH grant R01-DC009031.
Footnotes
Acknowledgments
We thank the staff of the Center for Balance Disorders, Baylor College of Medicine, for their assistance.