Hydration sensor development for use on infants in underserved areas

Kieser, Eduard (2015-12)

Thesis (MEng)--Stellenbosch University, 2015.

Thesis

ENGLISH ABSTRACT: To treat dehydration in infants a good measure of dehydration level is required. Clinical hydration scales are the standard for assessing hydration upon first examination. These scales rely on the training of the physician and do not perform well when applied in underserved settings where the vast majority of infant mortalities due to dehydration occur. This study proposes three sensors for the non-invasive assessment of dehydration. All three sensors measure proven indicators of hydration (capillary refill time, skin recoil time and skin temperature difference) and the aim is to eventually allow untrained personnel to reliably measure these hydration markers with greater inter-rater reliability. A proof of concept prototype of each sensor was fabricated and the sensors were tested in two separate studies. The skin recoil time and temperature profile sensors seemed to perform fairly well when tested on infants, however, because of the small sample size of the infant study, no real conclusions could be drawn about the link between the measured markers and hydration state. A systematic method of finding a model, that combines the best predictors, was presented, however the model performed only marginally better than the best individual marker, most likely due to wanting orthogonality between the predictors. Skin recoil time, which is the marker that performed best in the clinical study, achieved a sensitivity and specificity of 0.8 and 0.84 respectively for detecting 5% dehydration, which is better than the expected performance of any single commonly accepted hydration marker. Interestingly, this combination of sensitivity and specificity was achieved at a binary threshold of 0.08 seconds. As mentioned previously, the data does not make it possible to draw solid conclusions about the performance of the device, however the presented evidence makes a strong case for continued research into the presented methods.

AFRIKAANSE OPSOMMING: Dit is belangrik om dehidrasie in kleinkinders akkuraat te meet om sodoende die regte behandeling toe te pas. Kliniese hidrasie skale is die standaard vir die bepaling van hidrasie met die eerste ondersoek. Hierdie skale vertrou op die opleiding en ondervinding van die dokter en is nie baie betroubaar wanneer dit gebruik word in onbediende areas waar die oorgrootte meerderheid van kindersterftes weens dehidrasie is. Hierdie navorsing stel drie sensors voor vir nie-indringende assessering van dehidrasie. Al drie sensors meet bewese indikators van dehidrasie (kapillêre hervultyd, vel-terugkrimping en veltemperatuur-verskil) en die doel is om uiteindelik onopgeleide personeel toe te laat om hierdie merkers betroubaar te meet. ‘n Bewys-van-konsep prototipe van elke van die sensors is vervaardig en getoets in twee verksillende studies. Die vel-terugkrimping en veltemperatuur-verskil sensors het heel goed gevaar toe dit op die kleinkinders getoets is, maar vanweë die klein toetsgroep kon geen ware afleidings gemaak word tussen die verwantskap van die gemete merkers en die hidrasietoestand nie. ‘n Sistematiese metode om ‘n model te vind wat die beste voorspellers kombineer is voorgestel, maar die model het egter slegs marginaal beter presteer as die beste individuele merker, heel moontlik vanweë ‘n tekort aan ortogonaliteit tussen die voorspellers. Vel-terugkrimping, die merker wat die beste presteer het in die kliniese studie, het onderskeidelik ‘n sensitiwiteit en spesifiekheid van 0.8 en 0.84 behaal vir die opsporing van 5% dehidrasie, wat beter is as die verwagte verrigting van enige van die ander algemeen aanvaarde hidrasiemerkers. Interessant genoeg is die kombinasie van sensitiwiteit en spesifiekheid behaal by ‘n binêre grens van 0.08 sekondes. Soos voorheen genoem is dit met die data nie moontlik om konkrete gevolgtrekkings te maak rakende die gedrag van die toestel nie, maar die voorgestelde bewyse maak ‘n sterk saak vir verdere navorsing oor die voorgestelde metodes.

Please refer to this item in SUNScholar by using the following persistent URL: http://hdl.handle.net/10019.1/98029
This item appears in the following collections: