Rates of Sensor Loss in Unattended Home Polysomnography: The Influence of Age, Gender, Obesity, and Sleep-Disordered Breathing

Transcription

1 RATES OF SENSOR LOSS IN UNATTENDED HOME POLYSOMNOGRAPHY Rates of Sensor Loss in Unattended Home Polysomnography: The Influence of Age, Gender, Obesity, and Sleep-Disordered Breathing Vishesh K. Kapur MD MPH, David M. Rapoport MD, Mark H. Sanders MD, Paul Enright MD, Joel Hill MS, Conrad Iber MD, Joanna Romaniuk MS Department of Medicine, University of Washington, Seattle, WA (VKK); Department of Medicine, New York University, New York, NY (DMR); Departments of Medicine and Anesthesiology, University of Pittsburgh, Pittsburgh, PA (MHS); Respiratory Sciences Center, University of Arizona, Tucson, AZ (PE); Department of Medicine, Johns Hopkins University, Baltimore, MD (JH); Department of Medicine, University of Minnesota, Minneapolis, MN (CI); Department of Pediatrics, Case Western Reserve University, Cleveland, OH (JR). *Participating Institutions are listed under Acknowledgments. Objectives: To evaluate study failure and sensor loss in unattended home polysomnography and their relationship to age, gender, obesity, and severity of sleep-disordered breathing (SDB). Design: A cross-sectional analysis of data gathered prospectively for the Sleep Heart Health Study (SHHS). Setting: Unattended polysomnography was performed in participants' homes by the staff of the sites that are involved in SHHS. Participants: 6,802 individuals who met the inclusion criteria (age >40 years, no history of treatment of sleep apnea, no tracheostomy, no current home oxygen therapy) for SHHS. Results: A total of 6802 participants had 7151 studies performed of 6802 initial studies (90.6%) were acceptable. Obesity was associated with a decreased likelihood of a successful initial study. After one or more attempts, 6440 participants (94.7%) had studies that were judged as acceptable. The mean duration of scorable signals for specific channels ranged from 5.7 to 6.8 hours. The magnitudes of the effects of age, gender, BMI, and RDI on specific signal durations were not clinically significant. Conclusion: Unattended home PSG as performed for SHHS was usually successful. Participant characteristics had very weak associations with duration of scorable signal. This study suggests that unattended home PSG, when performed with proper protocols and quality controls, has reasonable success rates and signal quality for the evaluation of SDB in clinical and research settings. Key words: Unattended; polysomnography; quality; success INTRODUCTION POLYSOMNOGRAPHY (PSG) HAS BEEN WIDELY USED TO STUDY SLEEP AND BREATHING IN CLIN- ICAL AND RESEARCH SETTINGS. It is known that a number of factors can interfere with PSG study quality, including the improper application of sensors and signal failure or artifact generated by the subject or environment. Unattended home PSG monitoring is a technology that is being used for clinical and research purposes. Unlike attended PSG where personnel are available to detect and rectify technical difficulties, unattended studies are susceptible to failure due to such problems. If repeat home studies or attended PSG are required as a consequence of inad- Accepted for publication April 1999 Address correspondence to: Vishesh Kapur MD, MPH, Harborview Medical Center, Box , 325 Ninth Avenue, Seattle, WA Tel: , Fax: , vkapur@u.washington.edu equate unattended home evaluations, the cost of patient care or research increases. Although available published data indicate that PSG can be performed with reasonable success in the home, there has not been an adequate assessment of study failure rates and signal quality across a broad spectrum of community dwelling individuals. 1,2,3 In addition, there are no data addressing the relationship between study quality and subject characteristics such as age, gender, body habitus, and sleep-disordered breathing (SDB) severity. If these characteristics have a clinically significant impact on PSG quality, they may also affect the ability of PSG to accurately measure the nature and magnitude of SDB in certain individuals and thus potentially affect sleep information obtained from epidemiological studies. Identification of specific problems related to this methodology will facilitate the focus of efforts to improve technology. The Sleep Heart Health Study (SHHS) is a multi-center SLEEP, Vol. 23, No. 5,

2 study of SDB and cardiovascular disease in which unattended home PSG monitoring was used to evaluate over 6000 participants. 4 In the present study, PSG data from SHHS were examined to assess study failure rates and signal quality. The influences of several risk factors (age, gender, obesity, and severity of SDB) on study failure and signal quality were investigated. The implications of these results on the assessment of sleep and breathing in unattended home studies are discussed. METHODS Study Population SHHS participants were recruited from ongoing cohort studies including: the Atherosclerosis Risk in Communities (ARIC) Study, the Cardiovascular Health Study (CHS), the Framingham Heart Study (FHS), the Strong Heart Study (SHS), the Tucson Epidemiologic Study of Obstructive Airways Disease, the Health and Environment Cohort Study, and three New York City cohorts recruited to assess the effect of psychosocial risk factors on cardiovascular disease. 4,5 The institutional review board of each participating institution approved the study protocol. A total of 6,802 individuals from these cohorts who met inclusion criteria (age >40 years, no history of treatment of sleep apnea, no tracheostomy, no current home oxygen therapy) had an unattended baseline PSG performed between December 1995 and February Polysomnography Sleep studies were performed with the Compumedics PS-2 system (Compumedics Pty. Ltd, Abbotsville, AU) with the following recording montage: C3/A1 and C4/A2 electroencephalograms (EEG), right and left electrooculograms (EOG), a bipolar submental electromyogram (EMG), nasal/oral airflow (detected by thermocouple (Protec, Woodenville, WA)), chest and abdominal movement (recorded by inductive plethysmography bands), oxyhemoglobin saturation by pulse oximetry (Nonin, Minneapolis, MN), electrocardiogram (ECG) (recorded by bipolar lead), body position (using a mercury gauge), and ambient light employing a light sensor secured to the recording garment. 5,6 Leg movements were not recorded. Techniques for sensor attachment were standardized for the study. Sensors were attached to participants in the home using water-soluble adhesives, electrical conducting gels, and combinations of gauze and tape. Participants wore a specially designed vest that had pockets and pouches used to secure wires and the head box. This vest allowed the participant some freedom of movement without becoming entangled in the lead wires. After hookup, signals were visualized and sensor positions were modified to improve signal quality when needed. Impedance values were checked and EEG, EOG, and EMG electrodes were replaced if impedance values exceeded 5 kohms. Technicians were certified by written and practical examination and were required to perform five studies of acceptable quality in five non-cohort subjects before performing studies independently. The quality performance of individual technicians was monitored. Before leaving the home, the technician checked that the subject understood all aspects of the study and knew how to move about while wearing the recording unit. Depending on participant preference, electrodes and sensors were removed in the morning by the participant or technician. Study Quality Table 1 Characteristics of subjects as a function of the number of studies needed for success. All studies were reviewed at the Reading Center for signal quality. Studies were considered failed if they lacked one or more of the following: four hours of oximetry data, one EEG signal of sufficient quality to distinguish sleep from wake, or four hours of contiguous data from either abdominal, chest or thermocouple sensors. In general, participants with failed studies were asked to undergo repeat studies. For the purposes of our analysis, the duration in hours (0 9) of scorable signal assigned to channels was used as an ordinal categorical measure of signal quality. Duration Group Successful Studies Mean Age in years* (SD) Proportion Male Mean BMI* (SD) Median RDI 1 1 of 1 Attempts (N=6161) (11.0) (5.3) 2 1 of > 2 Attempts (N=279) (10.8) (6.7) 3 All Attempts Unsuccessful (N=362) (11.1) (5.9) *p< 0.01 SLEEP, Vol. 23, No. 5,

3 of scorable signal was defined as the period of signal recording of sufficient quality to allow interpretation of physiologic data, (i.e., respiratory events could be determined from the abdominal, chest, and thermocouple signals; oximetry fluctuations varied in a physiological manner [without abrupt changes] and showed the presence of a "good" waveform as displayed by the NONIN output; sleep could be distinguished from wakefulness on EEG; eye movements could be detected by EOG; and EMG fluctuations could be discerned suggesting changes in muscle tone). For channels that had bilateral sensors (EEG and EOG), the duration of the longest scorable signal was used. Signal quality was assessed at approximately half-hour intervals within a study and the total duration of scorable signal was rounded to the nearest whole hour that was still less than the total study time. In addition, measures of overall (all the signals listed above), respiratory (chest, abdominal, thermocouple, and oximetry) and staging (EEG, EOG, and EMG) quality were calculated using an average of the durations of the signals that make up these categories. Scoring of Polysomnograms Following review of study quality, each PSG was assigned to a trained and certified scorer at the Reading Center for manual scoring of sleep and breathing. The Reading Center had rigorous quality assurance procedures 6. In addition, a formal scoring reliability study that quantified within and between-scorer variability was performed early in the study. 7 Sleep stages were scored according to Rechtshaffen and Kales criteria. 8 Sleep time was defined as the total duration in hours of staged sleep during the study. Apnea was defined as a complete or almost complete (<25% of the baseline) cessation of airflow; hypopnea was defined as a decrease below 70% of baseline on chest, abdominal, or thermocouple channels for at least 10 seconds duration. To calculate the RDI, total number of respiratory events (apneas and hypopneas) associated with at least 4% desaturation was divided by total sleep time. Study duration was defined as the number of hours of recorded study. Covariates Age, gender, BMI (body mass index), and RDI were used as covariates to examine study quality. Age at the time of sleep study was defined in years. BMI was defined as weight (kg)/height (m). 2 The height measurement was obtained from the parent studies and thus predated the sleep study by as much as several years. Weight in kilograms was measured using a portable scale on the night of the polysomnogram with the participant wearing light clothes. If weight from the night of the polysomnogram was unavailable, a weight measurement obtained from the parent studies was used. Statistical Analysis All analyses were performed using SPSS data analysis software. 9 Mean values of normally distributed covariates (age and BMI) were compared using ANOVA. Proportions of subjects who were men were compared using the chisquare test. Median values of RDI, which has a highly skewed distribution, were compared using the Kruskal- Wallis test. Multiple logistic regression was performed to determine the associations with study success on initial attempt. Age, BMI, gender as well as nine indicator variables for the ten study sites that performed the studies were included as independent variables. Multiple linear regression was performed to examine the relationship between signal duration adjusted for study duration, to age, gender, BMI, and RDI. Tables 3 is based on the coefficients in the following regression equation which relates minutes of scorable signal (adjusted for study duration) to age, gender, BMI, and RDI: Minutes of signal = β o +β 1 (Age)+β 2 (Gender)+β 3 (BMI)+β 4 (RDI)+β 5 (hours of study)+β 6 (1/hours asleep) +β 7-15 (study sites 1 9) Study duration was included as an independent variable to adjust the duration of scorable signal for the length of the study since it would be expected that longer study duration would be related to duration of scorable signal. Nine indicator variables that represent each of ten sites that performed the studies were included to adjust for differences related to study site. The independent variable 1/(sleep Table 2 Duration of adequate signal. Signal Mean duration in hours % of study duration (Standard Deviation) (Standard Deviation) Abdomen (1.97) (26.2) Thermocouple (1.75) (22.6) Chest (1.95) (25.6) Chin (2.05) (26.8) EEG (1.33) (14.7) EEG (1.51) (17.6) EOG-R (1.14) (11.5) EOG-L (1.16) (11.7) Oximeter (1.03) (8.3) Abdomen and Chest = respiratory effort sensors Chin = Chin EMG % of study duration= signal duration/study duration in hours SLEEP, Vol. 23, No. 5,

4 time) was included in the model to avoid the potential for spurious correlation between RDI (number of respiratory events/sleep time) and signal duration since sleep time and signal duration were correlated. The values for the coefficients of age (β1), gender (β2), BMI (β3), and RDI (β4) obtained from multiple regression analysis represent the predicted change in minutes of scorable signal for a specified unit change in that particular subject characteristic. To test whether the linear specifications of age and BMI were appropriate, additional regression analyses using indicator variables to specify quartiles for these variables were performed. The results of these additional analyses were similar to the results obtained with the linear specification of the variables. To avoid heterogeneity in how studies were performed and in participant experience with polysomnography, only successful studies from the first recording attempt were included in regression analyses. A stringent criteria for statistical significance (p<0.01) was applied to all analyses to minimize the chances of false positive results with multiple analyses. RESULTS Table 3 Change in minutes of signal duration for a given change in covariate. Successful Studies and Number of Attempts A total of 6802 participants had 7151 studies performed of 6802 initial studies (90.6%) were acceptable (94.7%) participants ultimately had successful studies after one or more attempts. A comparison of participants based on whether the first study was successful (group 1), 2 or more studies were needed for success (group 2), or all studies were unsuccessful (group 3) is shown in Table 1. It should be noted that the majority of members of group 3 had only one study attempt so this group was not necessarily more likely to have a failed study than group 2. There were statistically significant differences in age and BMI between the groups but the magnitudes of these differences were quite small. Multiple logistic regression analysis of study success on initial attempt showed that in a model that included age, BMI, gender, and individual study site as independent variables, only BMI and several study sites were significantly related to study failure on initial attempt. A single unit increase in BMI increased the odds of study failure on the initial attempt by a factor of Dependent Variable Age Gender BMI RDI (each 10 years (male =1) (each 1 units BMI (each 10 unit change=1) change =1) change in RDI=1) β1 β2 β3 β4 Overall *** 0.9 Respiratory -1.9* *** 1.8 ** Abdomen -6.2*** 7.4 * -1.7*** 3.3** Chest *** 1.7 Oximetry * Thermocouple *** 2.1 Staging *** -0.3* -0.2 Chin ** -0.7* -1.1 EEG * EOG * Overall = average duration of seven signals (Abdomen, Thermocouple, Chest, Oximetry, Chin, EEG, and EOG) Respiratory = average duration of four respiratory signals (Abdomen, Thermocouple, Chest, and Oximetry) Staging = average duration of 3 staging signals (Chin, EEG, and EOG) * p<0.05, ** p<0.01, *** p<0.001 SLEEP, Vol. 23, No. 5,

5 Duration of Interpretable Signal for Individual Sensors Table 2 presents the mean duration in hours of interpretable signal among subjects whose initial studies were successful (group 1). Abdomen, chest, thermocouple, and chin signals were interpretable on average for 5.67 to 5.84 hours, while EEG, EOG and oximetry signals were interpretable on average for 6.48 to 6.75 hours. Mean study duration was 7.25 hours. Mean sleep time was 5.98 hours. Relation between Study Quality and Covariates The relation between hours of scorable signal duration adjusted for study duration and age, gender, BMI, and RDI was explored for the 6161 subjects whose initial study was successful (group 1). Although there were statistically significant relationships with some of the covariates, they were of small magnitudes. Table 3 provides the relevant coefficients of regression analyses for overall, respiratory, staging, and individual respiratory and staging signals. Overall signal duration was shorter in more obese subjects. Respiratory signal duration was shorter in more obese subjects and longer in subjects with higher RDI. Staging signal duration was higher in women. Abdomen signal duration was longer in younger subjects and in subjects with higher RDI. Abdominal and chest signal duration were shorter in more obese subjects. Chin EMG signal duration was shorter in men. DISCUSSION Study Success Rates and Number of Attempts Unattended home PSG as performed for SHHS had reasonable success rates. The method was successful in obtaining at least four hours of adequate respiratory and EEG data in 90.6% of initial attempts. Among subjects whose initial study was satisfactory, EEG, EOG and oximetry signals were of acceptable quality for 89% 93% of study duration while abdominal, thermocouple, chest and chin signals were of acceptable quality for a somewhat lower percentage of study duration (79% 81%). It should be noted that these numbers might underestimate the actual percent of scorable signal, as signal duration was rounded to the nearest whole hour that was less than study duration. Our results indicate that with appropriate training and well-designed protocols, unattended PSG, can be performed with reasonable success and quality for the purposes of quantifying SDB. We are not aware of published data on the failure rate for attended in-laboratory PSG, but we believe it is significantly less than what was obtained in our study for unattended home PSG. In our clinical experience, failed studies due to technical issues are rare in the laboratory since a technologist can correct most of these problems during the study. The likelihood that only one study attempt was required SLEEP, Vol. 23, No. 5, for success was decreased by obesity. The magnitude of the effect of obesity on the likelihood of a failed initial study was small. For example, a five-unit increase in average BMI in our population would increase the percentage of study failures from 9.5% to 10.5%. The duration's of abdominal, chest and thermocouple signals were negatively correlated with obesity suggesting the compromise of these signals may play a role in increasing the likelihood of study failure in obese subjects. The Relationship of Study Quality to Covariates Relationships between study quality and age, gender, BMI, and RDI were of particular interest for the purposes of SHHS. If the ability to measure SDB were influenced appreciably by these cardiovascular risk factors or potential risk factors, it would raise concerns about differential measurement error affecting the observed relationships between SDB and outcomes such as cardiovascular disease. Statistically significant relationships between some of these factors and specific signal durations were found, although the magnitudes of these relationships were very small. The impact of these relationships on the interpretation of the data collected for SHHS is negligible. There was a negative correlation between BMI and overall signal duration that was attributable to a negative correlation of BMI with abdominal, chest, and thermocouple signal duration. The negative relationship with abdominal and chest signal duration may be due to the compromise of the fit and function of impedance plethysmography by changes in abdominal and chest morphology that are associated with obesity. Respiratory signal duration was positively associated with RDI. One explanation of this relationship is that participants with increased sleep-disordered breathing had more easily detectable events, which led to a longer duration of scorable abdominal signal. Increasing age did not significantly impact respiratory signal duration, though abdominal signal duration decreased. Increasing age may be associated with abdominal morphology that compromises the fit of abdominal effort belts or the belts may be applied less effectively in older subjects. Staging signal duration was lower in males predominantly because chin EMG signal duration was lower. It may be that there was less effective application of submental electrodes in males because of the presence of facial hair. In general, the magnitudes of the effects of subject characteristics on signal duration were small (see Table 3). For example, a one-unit increase in BMI resulted in a decrease of only 0.8 minutes average signal duration and a 1.1 minute decrease in respiratory signal duration. Males had on average 4.4 minutes shorter staging signal duration than females. A 10-unit change in RDI resulted in a 1.8 minute increase in respiratory signal duration. In addition,

6 the effects on measured RDI of changes in specific signal duration were small. A one-hour increase in respiratory signal duration resulted in 0.5 increase in RDI (value derived from linear regression with RDI as the dependent variable and age, gender, BMI, study duration, and respiratory signal duration as independent variables). The small magnitude of the relationships between these characteristics and study quality is reassuring, since it is unlikely that differential measurement error of RDI due to age, gender, or obesity will significantly alter the observed relationship between sleep-disordered breathing and cardiovascular events. Our findings support the contention that unattended home PSG can be performed with adequate signal quality and success rates for clinical use in the evaluation of sleep disordered breathing. Our data suggest that approximately one out of ten patients would require repeat PSG (either laboratory or home based) to obtain adequate data, which would add to the cost of evaluation. Care must be taken in applying our findings to the clinical arena. A clinical population may have characteristics that make them more prone to study failure (obesity or nocturnal symptoms such as nocturia and unrestful sleep). In addition, if the degree of technician training and monitoring were not comparable in the clinical setting, one might expect a lower success rate. Since our PSG did not measure leg movements, our results do not directly address whether unattended home PSG that includes sensors to detect leg movements would have adequate success rates or signal quality for the purposes of detecting leg movements. At least one previously published study of unattended home PSG that included bilateral anterior tibialis EMG sensors, found that each measured parameter was scorable in greater than 95% of all epochs. 3 In summary, the present data provide estimates of the duration and quality of signals to be expected during unattended home polysomnography in epidemiologic studies. Although there were concerns that host characteristics would influence study quality, and study quality could influence RDI estimates, the magnitude of such effects appeared quite small. ACKNOWLEDGMENTS Participating Institutions and SHHS Investigators Framingham, MA: Boston University: George T. O'Connor, Sanford H. Auerbach, Emelia J. Benjamin, Ralph B. D'Agostino, Rachel J. Givelber, Daniel J. Gottlieb, Philip A. Wolf; University of Wisconsin: Terry B. Young. Minneapolis, MN: University of Minnesota: Eyal Shahar, Conrad Iber, Mark W. Mahowald, Paul G. McGovern, Lori L. Vitelli. New York, NY: New York University: David M. SLEEP, Vol. 23, No. 5, Rapoport, Joyce A. Walsleben; Cornell University: Thomas G. Pickering, Gary D. James; State University of New York, Stonybrook: Joseph E. Schwartz; Columbia University (Harlem Hospital): Velvie A. Pogue, Charles K. Francis. Sacramento, CA/Pittsburgh, PA: University of California, Davis: John A. Robbins, William H. Bonekat; University of Pittsburgh: Anne B. Newman, Mark H. Sanders. Tucson, AZ/Strong Heart Study: University of Arizona: Stuart F. Quan, Michael D. Lebowitz, Paul L. Enright, Richard R. Bootzin, Anthony E. Camilli, Bruce M. Coull, Russell R. Dodge, Gordon A. Ewy, Steven R. Knoper, Linda S. Snyder; Medlantic Research Institute Phoenix Strong Heart: Barbara V. Howard; University of Oklahoma Oklahoma Strong Heart: Elisa T. Lee, J. L. Yeh; Missouri Breaks Research Institute Dakotas Strong Heart: Thomas K. Welty. Washington County, MD: The Johns Hopkins University: F. Javier Nieto, Jonathan M. Samet, Joel G. Hill, Alan R. Schwartz, Philip L. Smith, Moyses Szklo. Coordinating Center Seattle, WA: University of Washington: Patricia W. Wahl, Bonnie K. Lind, Vishesh K. Kapur, David S. Siscovick, Qing Yao. Sleep Reading Center Cleveland, OH: Case Western Reserve University: Susan Redline, Carl E. Rosenberg, Kingman P. Strohl. NHLBI Project Office Bethesda, MD: James P. Kiley, Richard R. Fabsitz. The Sleep Heart Health Study (SHHS) acknowledges the Atherosclerosis Risk in Communities Study (ARIC), the Cardiovascular Health Study (CHS), the Framingham Heart Study (FHS), the Cornell Worksite and Hypertension Studies, the Strong Heart Study (SHS), the Tucson Epidemiology Study of Airways Obstructive Diseases (TES), and the Tucson Health and Environment (H&E) Study for allowing their cohort members to be part of the SHHS. We thank the many participants in the SHHS cohort who generously dedicated their time to study participation. We also thank the technical staffs of the clinical sites for their enthusiastic collection of data. We sincerely thank the Reading Center staff for their many contributions to the development and implementation of the protocols for processing, coding, and scoring the sleep studies. We thank Compumedics for modifying equipment for use by the SHHS. REFERENCES 1. Ferber R, Millman R, Coppola M, et al. Portable recording in the assessment of obstructive sleep apnea. Sleep 1994;17(4): Drewes AM, Nielsen KD, Taagholt SJ, et al. Ambulatory polysomnography using a new programmable amplifier system with online digitization of data: technical and clinical findings. Sleep 1996; 19(4):

7 3. Fry JM, DiPhillipo MA, Curran K, Goldberg R, Baran AS. Full polysomnography in the home. Sleep 1998;21(6): Quan SF, Howard BV, Iber C, et al. The sleep heart health study: design, rationale and methods. Sleep 1997;20(12): Sleep heart health study manual of operations. Seattle, WA:SHHS Coordinating Center, Redline S, Sanders M, Quan SF, et al. Methods for obtaining and analyzing polysomnography data for a multicenter study. Sleep 1998;21(7): Whitney CW, Gottlieb DJ, Redline S, et al. Reliability of scoring respiratory disturbance indices and sleep staging. Sleep1998; 21(7): Rechtshaffen A, Kales A. A manual of standardized terminology, techniques, and scoring systems for sleep stages of human subjects. Washington, DC: US Government Printing Office, 1968 (NIH Publication No. 204). 9. SPSS for Windows : base system user's guide, Release 6.0, SPSS Inc., SLEEP, Vol. 23, No. 5,