Zur Ermittlung des Energieverbrauches bei Mensch und Tier stehen verschiedene Methoden zur Verfügung. Im Jahre 1780 nutzte Lavoisier die Schmelzwassermenge, um den Energieverlust eines Meerschweinchens zu berechnen. Das Tier saß in einem von Eis umgebenen Kalorimeter, die von ihm abgegebene Wärme brachte das Eis zum Schmelzen. Derzeit sind die indirekte Kalorimetrie, die den Energieumsatz über den im Respirationsversuch gemessenen Gaswechsel von O2 und CO2 sowie die im Harn ausgeschiedene Stickstoffmenge bestimmt, und die Isotopendilutionsmethode, die mit der unterschiedlichen Ausscheidungsrate von markierten Wasserstoff- (2H) und Sauerstoff- (18O) Atomen im Urin arbeitet, der „Goldstandard“ für die Bestimmung des Energieverbrauchs. Seit einigen Jahren bis heute steht die Herzfrequenzmethode in der Diskussion. Sie nutzt die Beziehung zwischen Herzfrequenz und Sauerstoffverbrauch zur Ermittlung des Energieumsatzes. Alle genannten Methoden haben Vor- und Nachteile, insbesondere für den einfachen und schnellen täglichen Einsatz sowie bei Langzeitstudien. Deshalb werden Alternativen gesucht. Diese Dissertation untersucht die Beziehung zwischen der akzelerometrisch erfassten dreidimensionalen Körperbeschleunigung und der Herzfrequenz beim Pferd in verschiedenen Gangarten. Dabei wird die Herzfrequenz als Vergleichs- und Bezugsgröße verwendet. Sie stellt das direkte Bindeglied zum Sauerstoffverbrauch und damit Energieaufwand dar. Es wurden drei Versuchsvarianten durchgeführt. Die Pferde gingen an der Hand, „geführt“, liefen frei in einem umzäunten Oval, „freilaufend“, oder wurden „geritten“. Bei den beiden Varianten „geführt“ und „freilaufend“ kamen jeweils dieselben vier Pferde zum Einsatz, die Variante „geritten“ absolvierten fünf andere Tiere. Die Versuche folgten verschiedenen Schemata mit den Gangarten Schritt, Trab und, zum Teil, Galopp. Bei allen Versuchen wurden parallel die dreidimensionale Körperbeschleunigung mit einer Frequenz von 32 Hz sowie die Herzfrequenz gemessen. Die Pulsuhr speicherte im kleinstmöglichen Intervall von fünf Sekunden. Nach Aufbereitung der Beschleunigungsrohdaten wurde letztendlich der dynamische Anteil der dreidimensionalen Beschleunigung in Form von „fünf-Sekunden-Mittelwerten“ berechnet. Anschließend wurden diese Beschleunigungswerte über die Regressionsanalyse mit den Originalwerten der Herzfrequenz in Beziehung gesetzt. Dabei wurden die Übergangsphasen zwischen den Gangarten ausgenommen, da die beiden Parameter hier ein sehr unterschiedliches und zeitversetztes Verhalten zeigen. Bei der Analyse der Gangarten Schritt und Trab konnte gut mit dem Modell der einfachen linearen Regression (y = a + bx) gearbeitet werden, mit Hinzukommen der dritten Gangart, Galopp, erwies sich das Modell der polynomialen Regression (y = a + bx + cx²) von Vorteil. Die Stärke des Zusammenhanges der beiden Größen wurde durch den Korrelationskoeffizienten r angezeigt. Bei differenzierter Betrachtung der Versuchsvarianten und der einzelnen Pferde erreichte r Werte von 0,86 bis 0,94, bei zusammenfassender Betrachtung aller Pferde einer Versuchsvariante Werte zwischen 0,82 und 0,87, stets bei signifikanter Korrelation (p < 0,05). Somit kann für die Parameter Herzfrequenz und Beschleunigung ein signifikanter und starker Zusammenhang beschrieben werden. Sie verhalten sich dabei nicht proportional zueinander. Schlussfolgernd lässt sich sagen, dass die Akzelerometrie für bestimmte Zielstellungen und unter bestimmten Voraussetzungen eine geeignete Methode ist, um den Energieaufwand von Pferden zu bestimmen. Sie ist schnell und meist störungsfrei durchzuführen und im Gegensatz zur Herzfrequenz nahezu unabhängig von emotionalen Einflüssen. Des Weiteren bietet die Akzelerometrie die Möglichkeit, die Ermittlung des Energieumsatzes mit einer Verhaltensanalyse zu kombinieren. Bedingungen für ihren Einsatz sind eine situationsspezifische und möglichst individuelle Kalibrierung, denn die Beschleunigungsmessung weist insofern Nachteile auf, als dass sie die Auswirkungen von zum Beispiel Bodenbeschaffenheit, Umwelteinflüssen oder das Tragen einer Last auf den Energieumsatz nicht berücksichtigt. Die parallele Erfassung von Herzfrequenz und Beschleunigung kann zum Beispiel zur Analyse und Kontrolle von Trainingserfolgen genutzt werden. Somit bringt die Kombination von Herzfrequenz- und Beschleunigungsmessung klare Vorteile. / There are different opportunities to determine the consumption of energy in humans and animals. In 1780 Lavoisier used the quantity of melt water to calculate the energy loss of a guinea pig. The guinea pig was located inside a calorimeter which was surrounded by ice. The emitted heat induced the melting of the ice. At present both, indirect calorimetry that estimates energy expenditure from respiratory measurements of oxygen consumption and carbon dioxide production plus the excretion of nitrogen with the urine and the DLW-method that uses the different urinary elimination rates of the isotopes 2H and 18O are the so called “golden standard” for the calculation of energy consumption. For several years until now there has been a discussion about the heart rate-method. This method uses the correlation between heart rate and oxygen consumption for the calculation of energy expenditure. All above mentioned methods have pros and cons, especially for simple and quick every day application and for long-term studies. Therefore alternatives are searched. This dissertation examines the relation between the accelerometricly measured three-dimensional body acceleration and the heart rate in horses at different gaits. The heart rate has been used for comparison and as a reference item. It directly relates the acceleration with the oxygen consumption and thus with the energy expenditure. There have been three variants of trials. Horses were led by the hand (HD), moved freely (MF) in an enclosed oval or were ridden (R). In the HD- and MF-trials the same four horses were used, for the R-trials five other horses came into action. The trials followed different schemes with the gaits of walk, trot and gallop. At every trial three-dimensional body-acceleration with a logging frequency of 32 Hz and heart rate were measured simultaneously. The heart rate meter stored the heart rate in the smallest possible intervals of five seconds. After processing the crude data the dynamic part of the three-dimensional acceleration was calculated in form of “five-second-means”. After that the regression analysis was used to relate these acceleration data to the original heart rate data. In this process the transitional phases between the gaits were excluded because there both parameters have a highly varying and time-shifted relation. The model of simple linear regression (y = a + bx) suited well for analysing walking and trotting. With adding the third gait gallop the model of polynomial regression (y = a + bx + cx²) became more favourable. The correlation coefficient r showed the strength of the correlation between both parameters. By the separate inspection of the variants of trials and the individual horses r reached values from 0,86 to 0,94; pooling all horses of each variant of trials yields r-values from 0,82 to 0,87, always with a significant correlation (p < 0,05). Hence a significant and strong correlation can be attributed to the parameters heart rate and acceleration. They are not proportional to each other. In conclusion one can say: for specific aims and under certain conditions the accelerometry is an appropriate method to assess energy expenditure in horses. You can implement it quickly and mostly disturbance-free and in contrast to the heart rate it is nearly independent of emotional influence. Furthermore accelerometry gives the opportunity to combine the determination of the energy expenditure with the analysis of behaviour. A possibly individual and situation-specific calibration are the preconditions for its application. A setback of the accelerometry is that the effects of such factors like the condition of the ground, environmental influences or carrying weights are not taken into consideration. Simultaneous measurement of heart rate and body-acceleration can for example be used for analysing and controlling the success of training. Consequently there are clear advantages of combining the measurement of heart rate and acceleration.

Being able to effectively match energy intake to energy expenditure (EE) is an important aspect in preventing weight re-gain in the post-obese. Although it is generally agreed upon that resting EE decreases concomitantly with weight loss, there is no set standard comparing the deviations with differing weight loss protocols and additionally, controversy remains as to whether this decrease is greater than can predicted. In order to address these issues 2977 subjects were analyzed using a systematic review and the differences of both the protocol and length of various interventions in addition to sex were compared. Next, data was selected from this systematic review and 815 subjects were analyzed for weight loss-induced changes in resting EE, FM and FFM. Another subgroup of studies (n = 1450) was analyzed and compared against the Harris-Benedict prediction equation to determine whether the changes in resting EE were greater than what was expected. Finally, in order to determine which factors may be involved in regulating changes in resting EE during weight loss, a secondary analysis was performed on 28 post-menopausal women (age= 50.4 ± 2.0 yrs; BMI= 32.4 ± 5.2 kg/m²) who were submitted to a 6-month caloric restriction. Body composition (DXA), resting EE (indirect calorimetry), physical activity EE (PAEE) and total EE (TEE) (doubly-labelled water) were measured before and after the 6 month weight loss. Blood samples were collected before and after to measure leptin and peptide YY. The results indicate that there was indeed a depression in resting EE during weight loss regardless of the type of intervention utilized. Furthermore, these findings suggest that the changes could not fully be explained by changes of FM and FFM alone and that leptin may be an important contributor to the changes of resting EE during weight loss.

This study explores healthcare coverage disparity using a quantitative analysis on a large dataset from the United States. One of the objectives is to build supervised models including decision tree and neural network to study the efficient factors in healthcare coverage. We also discover groups of people with health coverage problems and inconsistencies by employing unsupervised modeling including K-Means clustering algorithm. Our modeling is based on the dataset retrieved from Medical Expenditure Panel Survey with 98,175 records in the original dataset. After pre-processing the data, including binning, cleaning, dealing with missing values, and balancing, it contains 26,932 records and 23 variables. We build 50 classification models in IBM SPSS Modeler employing decision tree and neural networks. The accuracy of the models varies between 76% and 81%. The models can predict the healthcare coverage for a new sample based on its significant attributes. We demonstrate that the decision tree models provide higher accuracy that the models based on neural networks. Also, having extensively analyzed the results, we discover the most efficient factors in healthcare coverage to be: access to care, age, poverty level of family, and race/ethnicity.

In this research, we apply data mining techniques to a nationally-representative expenditure data from the US to predict very high-cost patients in the top 5 cost percentiles, among the general population. Samples are derived from the Medical Expenditure Panel Survey’s Household Component data for 2006-2008 including 98,175 records. After pre-processing, partitioning and balancing the data, the final MEPS dataset with 31,704 records is modeled by Decision Trees (including C5.0 and CHAID), Neural Networks. Multiple predictive models are built and their performances are analyzed using various measures including correctness accuracy, G-mean, and Area under ROC Curve. We conclude that the CHAID tree returns the best G-mean and AUC measures for top performing predictive models ranging from 76% to 85%, and 0.812 to 0.942 units, respectively. Among a primary set of 66 attributes, the best predictors to estimate the top 5% high-cost population include individual’s overall health perception, history of blood cholesterol check, history of physical/sensory/mental limitations, age, and history of colonic prevention measures. It is worthy to note that we do not consider number of visits to care providers as a predictor since it has a high correlation with the expenditure, and does not offer a new insight to the data (i.e. it is a trivial predictor). We predict high-cost patients without knowing how many times the patient was visited by doctors or hospitalized. Consequently, the results from this study can be used by policy makers, health planners, and insurers to plan and improve delivery of health services.

The stride interval of typical human gait is correlated over thousands of strides. This statistical persistence diminishes with age, disease, and pace-constrained walking. Considering the widespread use of treadmills in rehabilitation and research, it is important to understand the effect of this speed-constrained locomotor modality on stride interval dynamics. To this end, and given that the dynamics of children have been largely unexplored, this study investigated the impact of treadmill walking, both with and without handrail use, on paediatric stride interval dynamics. An initial stationarity analysis of stride interval time series identified both non-stationary and stationary signals during all walking conditions. Subsequent scaling analysis revealed diminished stride interval persistence during unsupported treadmill walking compared to overground walking. Finally, while the correlation between stride interval dynamics and gross energy expenditure was investigated in an effort to elucidate the clinical meaning of persistence, no simple linear correlation was found.

Technological advances in physical activity measurement have increased the development and utilization of accelerometers and pedometers for assessing physical activity in controlled and free-living conditions. Individual differences in leg length, stride length and stride frequency may affect the reliability and validity of accelerometers in estimating energy expenditure. To address this theory, this thesis investigated the influence of leg length, stride length and stride frequency on accelerometer counts and energy expenditure using four accelerometers (AMP, Actical, MTI, and RT3) and one pedometer (Yamax). Eighty-six participants, age 8 to 40 (17.6 ± 8.0) years performed three ten-minute bouts of treadmill activity at self-selected speeds (4 to 12 km/h). Energy expenditure (kcal/min) was measured through expired gas analysis and used as the criterion standard to compare physical activity data from activity monitors. A 3 (models) x 2 (duplicates of each model) x 3 (speeds) x 7 (minutes) repeated measures ANOVA was used to assess intra-device, inter-device, and inter-model reliability. Coefficients of variation were calculated to compare within-device variation and between-device variation in accelerometer counts. Differences between measured and predicted energy expenditure were assessed across five height categories to determine the influence of leg length on the validity of accelerometer/pedometer data. Regression equations for each model were developed using mean activity counts/steps generated for each speed, adjusting for various predictor variables (i.e., age, weight, leg length). These were compared to model-specific equations to determine whether the addition of certain variables might explain more variance in energy expenditure. Leg length and stride frequency directly influenced variability in accelerometer data and thus predicted energy expenditure. At high speeds and stride frequencies counts began to level off in the Actical, however this did not occur in the other devices. Intra-device and inter-device variation in accelerometer counts was less than 10% and was lowest at very high speeds for the Actical, MTI, and RT3 (p<0.05). When compared to measured values, energy expenditure was consistently underestimated by the AMP, Actical, and Yamax models and consistently overestimated by the RT3 across speed. The MTI underestimated and overestimated energy expenditure depending on speed. Energy expenditure was both underestimated and overestimated to the greatest extent during the treadmill run for the tallest participants (p<0.05). Accelerometer counts or pedometer steps, when entered into regression equations with age, weight and leg length, explained from 85 to 94 % of the variance in measured energy expenditure, supporting the inclusion of these variables within manufacturer-based equations. These results suggest that individual differences in leg length and stride frequency affect the reliability and validity of accelerometer data and therefore must be controlled for when using accelerometry to predict energy expenditure.

The stride interval of typical human gait is correlated over thousands of strides. This statistical persistence diminishes with age, disease, and pace-constrained walking. Considering the widespread use of treadmills in rehabilitation and research, it is important to understand the effect of this speed-constrained locomotor modality on stride interval dynamics. To this end, and given that the dynamics of children have been largely unexplored, this study investigated the impact of treadmill walking, both with and without handrail use, on paediatric stride interval dynamics. An initial stationarity analysis of stride interval time series identified both non-stationary and stationary signals during all walking conditions. Subsequent scaling analysis revealed diminished stride interval persistence during unsupported treadmill walking compared to overground walking. Finally, while the correlation between stride interval dynamics and gross energy expenditure was investigated in an effort to elucidate the clinical meaning of persistence, no simple linear correlation was found.

Being able to effectively match energy intake to energy expenditure (EE) is an important aspect in preventing weight re-gain in the post-obese. Although it is generally agreed upon that resting EE decreases concomitantly with weight loss, there is no set standard comparing the deviations with differing weight loss protocols and additionally, controversy remains as to whether this decrease is greater than can predicted. In order to address these issues 2977 subjects were analyzed using a systematic review and the differences of both the protocol and length of various interventions in addition to sex were compared. Next, data was selected from this systematic review and 815 subjects were analyzed for weight loss-induced changes in resting EE, FM and FFM. Another subgroup of studies (n = 1450) was analyzed and compared against the Harris-Benedict prediction equation to determine whether the changes in resting EE were greater than what was expected. Finally, in order to determine which factors may be involved in regulating changes in resting EE during weight loss, a secondary analysis was performed on 28 post-menopausal women (age= 50.4 ± 2.0 yrs; BMI= 32.4 ± 5.2 kg/m²) who were submitted to a 6-month caloric restriction. Body composition (DXA), resting EE (indirect calorimetry), physical activity EE (PAEE) and total EE (TEE) (doubly-labelled water) were measured before and after the 6 month weight loss. Blood samples were collected before and after to measure leptin and peptide YY. The results indicate that there was indeed a depression in resting EE during weight loss regardless of the type of intervention utilized. Furthermore, these findings suggest that the changes could not fully be explained by changes of FM and FFM alone and that leptin may be an important contributor to the changes of resting EE during weight loss.

This study explores healthcare coverage disparity using a quantitative analysis on a large dataset from the United States. One of the objectives is to build supervised models including decision tree and neural network to study the efficient factors in healthcare coverage. We also discover groups of people with health coverage problems and inconsistencies by employing unsupervised modeling including K-Means clustering algorithm. Our modeling is based on the dataset retrieved from Medical Expenditure Panel Survey with 98,175 records in the original dataset. After pre-processing the data, including binning, cleaning, dealing with missing values, and balancing, it contains 26,932 records and 23 variables. We build 50 classification models in IBM SPSS Modeler employing decision tree and neural networks. The accuracy of the models varies between 76% and 81%. The models can predict the healthcare coverage for a new sample based on its significant attributes. We demonstrate that the decision tree models provide higher accuracy that the models based on neural networks. Also, having extensively analyzed the results, we discover the most efficient factors in healthcare coverage to be: access to care, age, poverty level of family, and race/ethnicity.

Technological advances in physical activity measurement have increased the development and utilization of accelerometers and pedometers for assessing physical activity in controlled and free-living conditions. Individual differences in leg length, stride length and stride frequency may affect the reliability and validity of accelerometers in estimating energy expenditure. To address this theory, this thesis investigated the influence of leg length, stride length and stride frequency on accelerometer counts and energy expenditure using four accelerometers (AMP, Actical, MTI, and RT3) and one pedometer (Yamax). Eighty-six participants, age 8 to 40 (17.6 ± 8.0) years performed three ten-minute bouts of treadmill activity at self-selected speeds (4 to 12 km/h). Energy expenditure (kcal/min) was measured through expired gas analysis and used as the criterion standard to compare physical activity data from activity monitors. A 3 (models) x 2 (duplicates of each model) x 3 (speeds) x 7 (minutes) repeated measures ANOVA was used to assess intra-device, inter-device, and inter-model reliability. Coefficients of variation were calculated to compare within-device variation and between-device variation in accelerometer counts. Differences between measured and predicted energy expenditure were assessed across five height categories to determine the influence of leg length on the validity of accelerometer/pedometer data. Regression equations for each model were developed using mean activity counts/steps generated for each speed, adjusting for various predictor variables (i.e., age, weight, leg length). These were compared to model-specific equations to determine whether the addition of certain variables might explain more variance in energy expenditure. Leg length and stride frequency directly influenced variability in accelerometer data and thus predicted energy expenditure. At high speeds and stride frequencies counts began to level off in the Actical, however this did not occur in the other devices. Intra-device and inter-device variation in accelerometer counts was less than 10% and was lowest at very high speeds for the Actical, MTI, and RT3 (p<0.05). When compared to measured values, energy expenditure was consistently underestimated by the AMP, Actical, and Yamax models and consistently overestimated by the RT3 across speed. The MTI underestimated and overestimated energy expenditure depending on speed. Energy expenditure was both underestimated and overestimated to the greatest extent during the treadmill run for the tallest participants (p<0.05). Accelerometer counts or pedometer steps, when entered into regression equations with age, weight and leg length, explained from 85 to 94 % of the variance in measured energy expenditure, supporting the inclusion of these variables within manufacturer-based equations. These results suggest that individual differences in leg length and stride frequency affect the reliability and validity of accelerometer data and therefore must be controlled for when using accelerometry to predict energy expenditure.