Calorimetric Validation of the Caltrac Accelerometer During Level Walking

Transcription

1 Calorimetric Validation of the Caltrac Accelerometer During Level Walking The primary purpoe of thi tudy wa to compare the Caltrac accelerometer output with meaured energy expenditure (Ee). Twenty-five volunteer (10 men, 15 women) walked on a level motor-driven treadmill at four different peed (54, 81, 104, and 130 m. min -1 ) with the Caltrac device affixed to the waitline. Each of the four experimental trial lated eight minute, and the teting wa completed within an hour. During the tet, oxygen conumption (Vo 2 ) (in L. min -1 and in ml. kg -1. min -1 ) and nonprotein repiratory exchange ratio were monitored by the Beckman Horizon metabolic cart. The accelerometer output at the end of each exercie bout wa alo monitored and ubequently divided by 8 to convert the reading to countṡmin -1. The mean Vo 2 (L. min -1 ) at teady tate (ie, 6th- 8th minute of exercie) wa converted to a caloric value. We obtained a moderate correlation coefficient (r) of. 76 between the accelerometer output and the Vo 2 (ml. kg -1. min -1 ) and a high correlation coefficient of.92 between the Ee and the accelerometer reading. The Caltrac accelerometer output (countṡmin -1 ) wa ignificantly higher (p <.01) than the Ee (kcal. min -1 ) at the four walking peed. The difference between the accelerometer output and the Ee rangedfrom 133% to 52.9%. The data were further analyzed with linear, polynomial, multiple, and tepwie regreion model. The reult of the analye revealed that the Caltrac accelerometer output i a valid predictor of Ee during level walking when the appropriate regreion equation i ued to adjut the value. Becaue the accelerometer device tend to overetimate Ee, the raw accelerometer reading hould be applied with caution. [Balogun J A, Martin DA, Clendenin MA Calorimetric validation of the Caltrac accelerometer during level walking. Phy Ther 69: , 1989] Joeph A Balogun Daniel A Martin Martha A Clendenin Key Word: Calorimetry; Energy expenditure; Equipment, exercie; Tet and meaurement, general. J Balogun, PhD, PT, i Senior Lecturer, Department of Medical Rehabilitation, Faculty of Health Science, Obafemi Awolowo Univerity, Ile-Ife, Oyo State, Nigeria, Wet Africa. He wa Viiting Reearch Scholar, Department of Phyical Therapy, Univerity of Florida, when thi tudy wa completed. D Martin, PhD, i Aitant Profeor, Department of Phyical Therapy, Health Science Center, PO Box J-154, Univerity of Florida, Gaineville, FL (USA). Addre correpondence to Dr Martin. M Clendenin, PhD, PT, i Profeor and Chairman, Department of Phyical Therapy, Health Science Center, Univerity of Florida. Thi reearch wa upported by a grant from the Diviion of Sponored Reearch, Univerity of Florida. Thi article wa ubmitted September 20, 1988; wa with the author for reviion for 10 week; and wa accepted February 23, Over the lat 20 year, there ha been a coniderable interet in the aement of phyical activity level and how activity level relate to cardiovacular fitne and health. 1-3 A variety of method have been ued in meauring phyical activity level including elf-report by quetionnaire and interview, direct obervation of phyical activitie, monitoring of heart rate by telemetry, calorimetric meaurement of oxygen uptake (Vo 2 ), iotope ratio ma pectrometry, and the ue of motion activity enor. Of all of thee method, the motion enor are reaonably nonobtructive, have the advantage of being objective, and are the mot cot-effective for clinical ue. 3-6 Phyical Therapy/Volume 69, Number 6/June /85

2 In recent year, many different type of mechanical and electronic motion enor have been manufactured and ued in epidemiological tudie to etimate phyical activity level. 3-6 The Caltrac accelerometer* i currently receiving increaed cientific interet becaue of the manufacturer' claim that it i a valid predictor of caloric expenditure, or energy expenditure (Ee), for a variety of phyical activitie. 7-9 The Caltrac device i a ingle-plane accelerometer that i deigned to monitor the up (acceleration)-and-down (deceleration) movement of the body. The movement i converted into electrical ignal and diplayed a digital readout repreenting a level of phyical activity. 7,8 The Caltrac accelerometer can accumulate up to 19,999 count, which i of ufficient capacity to lat for approximately 24 hour. 8 The device ue two lithium coin cell batterie, which can lat about 1,000 hour of continuou ue. A major limitation of the Caltrac accelerometer i that the device doe not repond to iometric (tatic) form of exercie and activitie in which the body weight i partially upported, a in bicycling or rowing. 9 Becaue of it light weight and low cot, the Caltrac accelerometer ha the potential to be ueful in phyical therapy practice to quantify the level of activity at work or during leiure activitie. 10 More pecifically, it can be ued to monitor patient' Ee during aerobic phyical activitie. It may be ueful in bariatric, cardiopulmonary rehabilitation, and pain management program. Some kepticim, however, ha been expreed concerning the validity of the Caltrac accelerometer in meauring Ee. 11 Thi tudy wa deigned to compare the Caltrac accelerometer output with meaured Ee. Review of the Literature Previou tudie have hown that the Caltrac accelerometer i reliable 9,10,12 and enitive to change in walking peed. 8-10,12 To date, however, few *Hemokinetic, Inc, 2923 Omunden Rd, Madion, WI tudie have evaluated the validity of the Caltrac accelerometer. Wong and co-worker validated the Caltrac accelerometer output againt the calorimetric meaurement of Vo 2. 7 Fifteen ubject walked or ran on a level, motor-driven treadmill for three minute each at 2, 3, 4, 6, and 8 mph. They reported "a roughly linear increae" in accelerometer reading with increaed Vo 2, but the trength of the relationhip wa not documented. In another tudy, Servai et al found a conitent aociation between daily Caltrac accelerometer output and the level of phyical activity. 8 The Ee in the tudy wa not meaured but intead derived from energy-cot table. Thee table repreent only etimate of Ee and are ubject to coniderable error. 9 In 1983, Montoye and aociate alo validated the Caltrac accelerometer reading againt the criterion of Vo 2. 9 The ubject (N = 21) performed 14 different phyical activitie, each lating four minute. The activitie involved 1) walking or running on the treadmill at different peed (2, 4, and 6 mph) and grade (0%, 6%, and 12%), 2) half-knee bend (28 and 48 bend. min -1 ), and 3) floor touche while bending at the knee (24 and 36 touche. min -1 ). They found that the Caltrac accelerometer output did not reflect the increaed Ee when the grade wa increaed during the walking or running activitie. The combined (pooled) data for all the activitie revealed a moderate correlation of.74, but a wide tandard error of etimate (SEE) of 6.6 ml O 2. kg -1. min -1. They derived linear regreion equation to predict Vo 2 (ml.kg -1. min -1 ) from the accelerometer output. Their regreion equation i of limited application in the clinical etting becaue patient following an exercie program are uually more intereted in their Ee than the amount of oxygen they conume. Therefore, it i of clinical relevance to derive a regreion equation to predict Ee (kcal. min -1 ) from the accelerometer output. In a recent report, Montoye et al found that the Caltrac accelerometer during bench-tepping and half-kneebending activitie etimate Ee "about a well a the force plate meaurement." 13 The correlation coefficient between the Caltrac accelerometer reading and the Vo 2 meaurement wa not reported, but the catter diagram plotted indicated a weak relationhip. In 1985, Klege and aociate validated the accelerometer reading againt the criterion of oberved phy ical activity behavior. 14 Fifty adult and 30 prechool children were oberved for one hour while engaged in freeplay activitie. The invetigator recorded the ubject' phyical activitie uing the Fargo Activity Time Sampling Survey Obervational Sytem. They reported moderate (r =.69) and low (r =.35) correlation coefficient between the Caltrac accelerometer output and the oberved phyical activity level for the adult and the children, repectively. They concluded that the Caltrac accelerometer may not adequately ae children' phyical activity level. In a follow-up tudy, Klege and Klege validated the Caltrac accelerometer output againt oberved phyical activity level of children in their natural environment over a nine-hour duration. 15 A moderate correlation (r =.54) wa obtained between the accelerometer output and the oberved phyical activity level. They attributed thi finding to the longer aement period ued, and concluded that the Caltrac accelerometer may be inappropriate for meauring the level of phyical activity of children. We found a total of ix validation tudie on the Caltrac accelerometer in the literature; four of the tudie ued Vo 2 a the criterion variable, and the remaining two tudie ued direct obervation a the criterion meaure. The direct obervation method i ubjective and too time-conuming for clinical ue. 3-6 In contrat, the calori- 86/502 Phyical Therapy/Volume 69, Number 6/June 1989

3 metric meaurement of Vo 2 i highly reproducible and a valid meaure of Ee. 16 New method of meauring phyical activity, therefore, are often validated againt Vo ,17 In laboratory and field tudie, Vo 2 during phyical activitie i meaured by the indirect calorimetric method. Thi method involve the collection of the expired ga uing a Dougla bag or Kofranyi-Michaeli (Max Planck) repirometer. 16 At teady tate, aliquot ample of the expired ga are withdrawn and analyzed for oxygen and carbon dioxide concentration, which are ubequently ued in the computation of Vo 2. Statement off the Problem From a clinical perpective, the validity of an intrument i indeed important. The validity contruct anwer the quetion, Doe the intrument meaure what i purport to meaure? In thi cae, doe the Caltrac accelerometer really meaure Ee? Although the reproducibility of the Caltrac accelerometer output i not in doubt, 9,10,13 the validity of the accelerometer in meauring Ee i unclear currently. 11 For the Caltrac accelerometer to have a wider clinical application, further tudie validating the device output againt the calorimetric method are warranted. The purpoe of thi tudy were to 1. Determine the relationhip between the Caltrac accelerometer output and the calorimetric meaurement of Vo 2 during level walking. 2. Determine whether difference exit between the Caltrac accelerometer output (count. min -1 ) and the meaured Ee (kcal.min -1 ) at different work load. 3. Derive regreion equation to predict Ee from the Caltrac accelerometer output, if difference were found between the Caltrac reading and the meaured Ee. We expected that 1) a highly ignificant correlation would exit between the Caltrac accelerometer reading and the Vo 2 meaurement, 2) no difference would exit between the Caltrac output and the meaured Ee, and 3) Ee can reliably be predicted (p <.05) from the accelerometer output. Method Subject Twenty-five healthy volunteer (10 men, 15 women), between the age of 18 and 38 year, from the Univerity of Florida participated in thi tudy. On the average, the ubject' age wa 24.7 ± 5.4 year, their height wa ±9.1 cm, and their body weight wa 67.4 ± 12.6 kg. Mot of the ubject were involved in recreational activitie, but none wa an elite athlete. The reearch protocol wa approved by the Intitutional Review Board of the Univerity of Florida. The medical hitory of all potential ubject wa reviewed, and thoe with cardiorepiratory and metabolic diorder were excluded from the tudy. Before teting, the ubject were informed of the teting procedure and rik of the tudy, and each ubject igned an informed conent form. Experimental Deign SenorMedic Corp, 1630 S State College Blvd, Anaheim, CA A quai-experimental protocol where each ubject acted a hi or her own control wa ued. 18 Specifically, each ubject walked on a level (0% grade), motor-driven treadmill at four different peed (54, 81, 104, and 130 m. min -1 ). Each of the four experimental trial lated eight minute, and all teting wa completed within one hour. The ubject' Vo 2 and accelerometer output were monitored during each trial. We elected thee walking peed becaue they repreent the range of peed adopted during mot recreational activitie, and we ued an eight-minute exercie duration to enure that the ubject (irrepective of their phyicalfitnetatu) attained a teady tate during the exercie bout. 19 We did not include graded walking in the experimental deign becaue previou tudie have hown that the Caltrac accelerometer tend to underetimate Ee during graded walking and running phyical activitie. 9,12 The room temperature during the teting wa maintained between 20 and 22 C. The relative humidity and barometric preure were maintained between 52% and 56% and 758 and 766 mm Hg, repectively. Procedure On arrival in the laboratory, the ubject' weight and height were meaured. Thee data and other peronal data uch a age and ex were programmed into the Caltrac accelerometer. The ubject were then introduced to the laboratory equipment and were allowed to practice walking on the treadmill until they became confident. Subequently, the Caltrac accelerometer wa affixed to the wait belt at the lumbar region, and the ubject were progreed to walking on the treadmill. The four experimental condition (ie, 54-, 81-, 104-, and 130-m. min -1 walking peed) were preented in equential order tarting with the 54-m. min -1 walking peed. Each of the teting condition lated eight minute, and adequate ret period of about five minute' duration were allowed between exercie bout. The teting wa reumed when the ubject re-attained their pre-exercie reting heart rate. During the tet, Vo 2 (in L. min -1 and in ml. kg -1. min -1 ) and nonprotein repiratory exchange ratio (RER) were monitored by the Beckman Horizon metabolic cart. 1 " For all the ubject, the metabolic cart' ga analyzer were calibrated againt tandard ga mixture hortly before the initial exercie bout and between the econd and the third exercie bout. The Vo 2 and RER were monitored every 30 econd, but only the value at teady tate (ie, 6th-8th minute of exercie) were ued for data analyi. The Caltrac accelerometer output Phyical Therapy/Volume 69, Number 6/June /87

4 wa monitored every two minute during the teting, but only the total reading at the end of the eighth minute were ued for data analyi. All ubject completed the four experimental trial; however, two of the ubject jogged intead of walked during the fatet peed (130 m. min -1 ). Computation The mean Vo 2 (L. min -1 and ml. kg -1. min -1 ) and RER from ixth through eighth minute of exercie wa computed. The Caltrac accelerometer output at the end of each exercie trial wa divided by 8 to convert the reading to count. min -1 The Vo 2 (L. min -1 ) wa converted to a caloric value (kcal. min -1 ) a decribed by McArdle et al 16 : Ee (kcal. min -1 ) = [Vo 2 (L. min -1 ) thermal equivalent of oxygen for RER] (1) The thermal equivalent of oxygen for the different RER were obtained from tandard nomogram. 16 Data Analyi We computed the Pearon productmoment correlation coefficient (r) for the combined (pooled) data to determine the relationhip between the accelerometer output and the meaured Vo 2 (ml.kg -1. min -1 ) and Ee (kcal. min -1 ). A paired t tet wa ued to determine ignificant difference between the Caltrac accelerometer output and the Ee at different walking peed. The paired t-tet tatitical procedure wa twotailed at an alpha level of.05. Both linear and polynomial regreion model were applied to etablih a predictive equation between the accelerometer output and the criteria variable (ie, Vo 2 and Ee). We determined the predictability of the regreion equation with the analyi of variance (ANOVA) procedure. Multiple and tepwie regreion model were applied to determine the contribution of age, body weight, and height to the prediction of the criteria variable. The tepwie regreion procedure elect the predictor variable in the order of their relative trength in predicting the criteria variable. The tolerance level wa et at a probability of.01 (F 4.00). Thee tatitical tet were performed on a Macintoh Plu micro-computer uing the StatView 512+ tatitical package. Reult Apple Computer, Inc, Mariani Ave, Cupertino, CA Brain Power, Inc, Ventura Blvd, Suite 250, Calabaa, CA Fig. 1 Scatter plot of meaured oxygen uptake and Caltrac accelerometer outpu howing 'bet line of fit." Figure 1 i a catter diagram illutrating the relationhip between the Caltrac accelerometer output and the meaured Vo 2 (ml. kg -1. min -1 ). A poitive correlation (r =.76, p <.001) wa found between the accelerometer reading and the Vo 2. The relationhip between the Caltrac accelerometer output and the meaured Ee i hown in Figure 2. The catter plot revealed a poitive linear relationhip (r =.91, p <.0001) between the accelerometer output and the Ee. The reult of the paired t tet ued to determine difference between the accelerometer output and the meaured Ee are ummarized in Table 1. The analyi revealed that the Caltrac accelerometer output wa ignificantly higher (p <.001) than the meaured Ee at the different walking peed. The difference between the Caltrac output and the meaured Ee ranged from 13.3% to 52.9%. Becaue the Caltrac accelerometer overetimated Ee (Fig. 3), it wa neceary to adjut for the meaurement error through the ue of a regreion equation. The reult of the regreion analye are preented below. 88/504 Phyical Therapy/Volume 69, Number 6/June 1989

5 The correlation coefficient, coefficient of determination (R 2 ), SEE, and F ratio of the ANOVA for the different regreion equation are ummarized in Table 2. The linear and polynomial regreion equation to predict Vo 2 (ml. kg -1. min -1 ) from the accelerometer output (X) are a follow: Vo 2 (ml. kg -1. min -1 ) = [X (count-min -1 )] (2) Vo 2 (ml. kg -1. min -1 ) = [X (count. min -1 )] [X 2 (count-min -1 )] (3) The ANOVA for both linear and polynomial regreion model (Tab. 2) revealed that Vo 2 (ml. kg -1. min -1 ) can reliably be predicted (p <.0001) from the accelerometer output. The quadratic equation (equation 3), however, wa a better (p <.01) predictor of Vo 2 than wa the linear equation (equation 2). The regreion curve for the quadratic equation i plotted in Figure 4. The linear and polynomial regreion equation decribing the relationhip between the meaured Ee and the accelerometer output (X) are a follow: Ee (kcal. min -1 ) = [X (count-min -1 )] (4) Ee (kcal. min -1 ) = [X (count. min -1 )] [X 2 (count-min -1 )] (5) The reult of the ANOVA for both model (Tab. 2) revealed that Ee can reliably be predicted (p <.0001) from the accelerometer output. The quadratic equation (equation 5) wa not a better predictor (p >.05) of Ee than wa the linear equation (equation 4). l in = 2.54 cm. Fig. 2. Scatter diagram of meaured energy expenditure and Caltrac accelerometer output howing "bet line of fit." Table 1. Comparion of Caltrac Accelerometer Output and Meaured Energy Expenditure (Ee) at Different Walking Speed Walking Speed (m. min -1 ) a p <.001. The addition of age, body weight, and height into an equation containing the accelerometer output (X) ignificantly increaed (p <.05) the accuracy of prediction of Vo 2 (Tab. 3). The multiple regreion equation i a follow: Accelerometer Output (count. min -1 ) Vo 2 (ml kg -1 min -1 ) = [X (count min -1 )] [age (yr)] [weight (kg)] [height (in) 1 ] (6) The reult of the tepwie regreion analyi for Vo 2 (ml. kg -1 min -1 ) are ummarized in Table 3. The analyi Meaured Ee (kcal. min -1 ) Difference (%) revealed that the bet predictor of Vo 2 were the accelerometer output and body weight. They accounted for 58.3% and 17.9% of the total variance (76.4%) in Vo 2, repectively. The multiple regreion equation for the meaured Ee i a follow: Ee (kcal. min -1 ) = [X (count. min -1 )] [age (yr)] [weight (kg)] [height (in)] (7) t P a Phyical Therapy/Volume 69, Number 6/June /89

6 The tepwie regreion analyi of the meaured Ee data revealed that the only viable predictor of Ee wa the accelerometer output. The contribution of age, body weight, and height toward the prediction of Ee wa minimal. The accelerometer output accounted for 83.8% of the total variance (84.2%) in Ee (Tab. 2). Dicuion Thi tudy wa deigned to compare the Caltrac accelerometer output with meaured Ee. We obtained a moderate correlation of.76 (p <.001) between the accelerometer reading and the meaured Vo 2 (ml. kg -1. min -1 ) and a high correlation of.92 (p <.0001) between the accelerometer output and the calculated Ee. Thee finding upported our reearch hypothei. The correlation coefficient that we obtained between the Vo 2 and the accelerometer output (r =.76) i conitent with the value (r =.74) reported by Montoye et al under imilar laboratory teting condition. 9 The SEE of 3.8 ml. kg -1. min -1 that we obtained for the linear regreion equation (equation 3), however, i much lower than the 6.6 ml. kg -1. min -1 reported by Montoye and aociate. 9 The higher SEE obtained in the tudy of Montoye et al may partially be attributed to the variety of phyical activitie included in their deign. Some of the activitie (ie, half-knee bending and floor touche while bending at the knee) are predominantly iometric form of exercie, which ingle-plane accelerometer cannot meaure Becaue previou validation tudie 7-9,13-15 did not derive regreion equation to predict Ee, further comparion cannot be made. The Caltrac operational manual indicate that the accelerometer "ha been cutomized to meaure the calorie your body need baed upon your weight, height, age, ex, caloric intake and activity level." 20 Baed on thi information, we did not expect to find a ignificant difference between the accelerometer output and the meaured Ee. Contrary to our expectation, we found that the Caltrac Fig. 3. Mean and tandard deviation for Caltrac accelerometer output and meaured energy expenditure. accelerometer ignificantly overetimated Ee (p <.01) during level walking (Tab. 1). We ued the Student' t tet tatitical procedure to determine whether there wa a ignificant difference in the accelerometer output and the meaured Vo 2 between the male and female ubject. The reult of the analyi Table 2. Regreion Analyi Linear regreion for Vo 2 d revealed that the accelerometer output (Tab. 4) and the energy expended (Tab. 5) by the male ubject were ignificantly higher (p <.001) than the female ubject at the different walking peed. An analyi of the ubject' phyical characteritic data revealed that the male ubject were heavier (t = 5.6, p <.0001), taller Analyi of Variance Summary for Different Regreion Analye Polynomial regreion for Vo 2 Linear regreion for Ee e Polynomial regreion for Ee Multiple regreion for Vo 2 Multiple regreion for Ee r a a Pearon product-moment correlation coefficient. b Standard error of etimate for the regreion equation. c p <.0001 d Vo 2 = oxygen conumption. e Ee = energy expenditure. R SEE b F c /506 Phyical Therapy/Volume 69, Number 6/June 1989

7 (t = 34,p <.01), and older (t = 3.5, p <.01) than the female ubject. The higher accelerometer and caloric output by the male ubject, therefore, may be attributed to their greater body ma. The reult of the Student' t tet for Vo 2, expreed a ml. kg -1. min -1 to eliminate the effect of body weight, howed no ignificant difference (p >.05) between the male and female ubject (Tab. 6), thu upporting our peculation. We ued different regreion model in an attempt to obtain the mot accurate regreion formula for the Vo 2 and Ee criteria variable. When everal model are ignificant, there are no univerally accepted rule for electing the bet regreion equation. 21 Apparently, clinician mut decide which equation bet meet their need taking into conideration the time and cot of meauring the predictor variable. For example, the addition of age, body weight, and height to the Vo 2 linear regreion equation containing the accelerometer output increaed the coefficient of determination from.583 to.764 (31%) and decreaed the SEE from 378 to 2.88 (24%) (Tab. 2). The eae with which peronal data uch a age, body weight, and height could be meaured and the ignificant increae (p <.05) in the predictive accuracy of the regreion upport the ue of the multiple regreion equation when an accurate etimate of Vo 2 i ought from the accelerometer output. We found no ignificant difference in the predictive accuracy between linear, polynomial, and multiple regreion formula for Ee. Becaue of it implicity, it will be more appropriate to ue the linear regreion formula to predict Ee from the accelerometer output. The reult of the tepwie regreion analye revealed that the bet predictor of Vo 2 were the accelerometer output and body weight. The accelerometer output wa the only viable predictor of Ee (kcal. min -1 ). Thee finding ugget that the Caltrac accelerometer output i a valid predictor of Ee during level walking Fig. 4. Scatter plot of meaured oxygen uptake and Caltrac accelerometer outpu howing regreion curve for quadratic equation. when the appropriate regreion equation i ued. Limitation of Study The regreion equation derived in thi tudy can be ued by adult between 18 and 38 year of age to predict their Ee while walking on a level terrain with the accelerometer device affixed to the lumbar region. The regreion equation i not valid while walking on a lope or during running activitie. The equation may not be applicable when the Caltrac device i affixed to other body location. In a recent tudy, Balogun et al obtained ignificantly higher (p <.05) accelerometer output when the Caltrac device wa worn at the waitline a compared with when it wa affixed to the chet line. 12 It i currently unclear whether the accelerometer output will be affected ignificantly if the device i affixed to a different location on the waitline. Thi problem warrant invetigation becaue ome client prefer to wear the Caltrac device at the back of the waitline, wherea other client prefer Table 3. Stepwie Regreion Analyi Summary for Oxygen Conumption Showi Change in Multiple Regreion Coefficient (r) and Coefficient of Determination (R 2 with Addition of Different Variable Stage of Analyi Accelerometer output (X) X and body weight All variable (X, age, body weight, and height) a p <.001. r R R Change F a Phyical Therapy/Volume 69, Number 6/June /91

8 to wear it at the ide or at the abdominal region of the waitline. We obtained a high correlation (r =.87) for the Vo 2 multiple regreion equation. The SEE for the regreion equation, however, wa 2.88 ml. kg -1. min -1. The SEE indicate that coniderable variability remain, depite the trong predictive power of the equation. Even though the SEE obtained in thi invetigation for the regreion equation (Tab. 2) are much lower than the 6.6 ml. kg -1. min -1 reported by the manufacturer of the Caltrac device, 9 the high SEE mut be appreciated by phyical therapit uing the accelerometer to predict Vo 2 during level walking. Clinical Implication The Caltrac accelerometer output can be ued by the phyical therapit to objectively monitor the patient' relative level of activity during aerobic exercie level-walking program. Becaue the Caltrac device tend to overetimate Ee (Fig. 3), the accelerometer output may create a fale impreion on the amount of calorie expended during phyical activitie. The Caltrac accelerometer output, therefore, hould be applied with caution in making clinical deciion on patient' Ee. Patient uing the Caltrac device hould be informed about the predictive accuracy of the accelerometer output. The phyical therapit may educate the patient on the poible ue of the regreion equation when an accurate etimate of the Ee i needed. It i neceary to apply the regreion equation in weight-lo program becaue the feedback provided by the accelerometer erve a an impetu for a poitive change in behavioral modification of the patient. Concluion Our tudy revealed a moderate correlation of.76 between the accelerometer output and the meaured Vo 2 (ml. kg -1. min -1 ) and a high correlation of.92 between the meaured Ee and the accelerometer reading. The Table 4. Comparion of Male and Female Subject' Caltrac Accelerometer Outp at Different Walking Speed Walking Speed (m. min -1 ) a p <.001. Caltrac accelerometer output wa ignificantly higher (p <.01) than the meaured Ee at the different walking peed. We found that the bet predictor of Vo 2 were the accelerometer output and body weight, although the accelerometer output wa the only Accelerometer Output (count. min -1 ) Male (n = 10) Female (n = 15) Table 5- Comparion of Male and Female Subject' Meaured Energy Expenditur (Ee) at Different Walking Speed Walking Speed (m. min -1 ) a p <.001. Ee (kcal. min -1 ) Male (n = 10) viable predictor of Ee. Our finding revealed that the Caltrac accelerometer output i a valid predictor of Ee during level walking when the appropriate regreion equation i ued. Becaue the accelerometer device tend to overetimate Ee, the raw Female (n = 15) Table 6. Comparion of Male and Female Subject' Meaured Oxygen Conumption (Vo 2 ) at Different Walking Speed Walking Speed (m. min -1 ) a p >.05. Vo 2 (ml. kg -1 min -1 ) Male (n = 10) Female (n = 15) t a t a t a /508 Phyical Therapy/Volume 69, Number 6/June 1989

9 accelerometer reading hould be ued with caution. Acknowledgment We thank Richard R Gutekunt, PhD, Dean, College of Health Related Profeion, Univerity of Florida, for hi aitance in proceing the reearch grant. Alo, we would like to acknowledge the cooperation of the phyical therapy tudent at the Univerity of Florida. Reference 1 LaPorte RE, Kuller LH, Kupfer DJ, et al: An objective meaure of phyical activity for epidemiological reearch. Am J Epidemiol 109: , Montoye HJ, Taylor HL: Meaurement of phyical activity in population tudie: A review. Human Biol 56: , Sari WHM: Habitual phyical activity in children: Methodology and finding in health and dieae. Med Sci Sport Exerc 18: , LaPorte RE, Black-Sander R, Cauley JA, et al: The epidemiology of phyical activity in children, college tudent, middle-aged men, menopaual female and monkey. J Chronic Di 75: , LaPorte RE, Adam LL, Savage DD, et al: The pectrum of phyical activity, cardiovacular dieae and health: An epidemiologic perpective. Am J Epidemiol 120: , LaPorte RE, Montoye HJ, Caperon CJ: Aement of phyical activity in epidemiologic reearch: Problem and propect. Public Health Rep 100: , Wong TC, Webter JG, Montoye HJ, et al: Portable accelerometer device for meauring human energy expenditure. IEEE Tran Biomed Eng 6: , Servai SB, Webter JG, Montoye HJ: Etimating human energy expenditure uing an accelerometer device. IEEE Frontier of Engineering in Health Care 8: , Montoye HJ, Wahburn R, Servai SB, et al: Etimation of energy expenditure by a portable accelerometer. Med Sci Sport Exerc 15: , Balogun JA, Farina NT, Fay E, et al: Energy cot determination uing a portable accelerometer. Phy Ther 66: , Smith LK: Commentary. Phy Ther 66: , Balogun JA, Amua LO, Onyewadume IU: Factor affecting Caltrac and Calcount accelerometer output. Phy Ther 68: , Montoye HJ, Servai SB, Webter JG: Etimation of energy expenditure from a force platform and an accelerometer. In: Proceeding of the Eighth Commonwealth and International Conference on Sport, Phyical Education, Dance, Recreation and Health. London, England, 1986, pp Klege RC, Klege LM, Swenon AM, et al: A validation of two motion enor in the prediction of child and adult phyical activity level. Am J Epidemiol 122: , Klege LM, Klege RC: The aement of children' phyical activity: A comparion of method. Med Sci Sport Exerc 19: , McArdle WD, Katch FI, Katch VL: Exercie Phyiology: Energy, Nutrition and Human Performance. Philadelphia, PA, Lea & Febiger, Schoeller DA, Ravuin E, Schutz Y, et al: Energy expenditure by doubly labeled water: Validation in human and propoed calculation. Am J Phyiol 250: , Campbell DT, Stanley JC: Experimental and Quai-Experimental Deign for Reearch. Boton, MA, Houghton Mifflin Co, Bubb WJ, Martin AD, Howley ET: Predicting oxygen uptake during level walking at peed of m/min. Journal of Cardiopulmonary Rehabilitation 5: , Caltrac : Peronal Activity Computer, Intruction Manual. Madion, WI, Caltronic Div of Hemokinetic, Inc, Martin D, Acker JE: Predicting aerobic capacity during the modified Naughton treadmill protocol in patient with coronary artery dieae. Journal of Cardiopulmonary Rehabilitation 8: , 1988 Phyical Therapy/Volume 69, Number 6/June /93