Accuracy and Precision Comparison of Temptraq

Design A repeated measures within-group comparative design will be used for this study.

Sample/Setting To address the specific aims of this study the investigators will identify patients (N = 40) who have a pulmonary artery (PA) catheter in place (patients in intensive care units). Letters of support are attached from the unit managers where the study will take place (see attached).

Procedure Once a patient is identified, the researcher will record basic demographic data (see data collection form attached) and the axillary area will be assessed for visible signs of any skin conditions. If there are no visible signs of a skin condition the researcher will apply the Temp-Traq thermometer to the right or left axilla of the subject and record the location of the placement. The researcher will record both the PA and Temp-Traq recordings of subject's temperature (taken at the same time) on the data collection form (see attachment) at four data points; baseline (5 minutes post application of Temp-Traq) and every two hours (+/- 15 minutes) after baseline X 3. Previous testing of the TempTraq® on children and healthy adults has not resulted in any adverse skin reactions. However, as this is a new thermometer format and there is not a significant body of literature on potential skin reactions, the investigators will assess skin before and after application. Thus, at the end of the last reading the Temp-Traq thermometer will be removed and the skin will be assessed for erythema. If erythema is present the patient's nurse and physician will be notified by the researcher.

Protection of Human Subjects Since non-invasive temperature monitoring poses no risk to patients and is part of usual care and since the investigators will not collect or record any protected health information (PHI), written informed consent will not be sought. Patients will be identified by their study ID number only. The data will be collected by a registered nurse data collector familiar with the intensive care setting and data will be stored in a locked file cabinet in the researchers locked office. Temperature data for the Temp-Traq is collected through the Temp-Traq Application (AP) which will be placed on a Cleveland Clinic Approved and encrypted i-pad. Once the study is completed and all data has been transferred to the study database, the Temp-Traq application and all of its data will be deleted from the i-pad.

Analysis The investigators will use the TOST (Two One Sided t-tests) method for equivalence testing. This method requires a defined range of mean differences between two test methods, an estimate of the precision of the measurement of the two systems, and an estimate of the size of the possible difference in the means of the two methods under consideration. The TOST null hypothesis is a joint null hypothesis that the mean measurement differences between the two methods is greater than a critical lower bound and less than a critical upper bound. If the null is rejected then the investigators can conclude that the absolute difference of the means for the two groups falls within the specified range. It was determined that to be considered equivalent with respect to accuracy the mean measurements of the two methods should be within ± 0.2 degrees of each other. Precision tests for Temp-Traq thermometer using accepted ASTM test methods for various combinations of temperature and humidity provided a range of measurement variation between .0000435 and .019928 with a mean of 0.000254500. If the investigators assume both test methods exhibit the same levels of precision then the variance of the differences between their two means will reduce to two times the values listed previously. This will result in estimates of the standard deviations of the mean differences of .0093, .032, and .063 respectively.

The two methods for body temperature that will be examined are core temperature measurements with a PA catheter and temperature measurements using the Temp-Traq (measured in Celsius). Paired temperature measurements for both methods will be taken simultaneously every 2 hours for a 6 hour period. This will provide 4 repeated temperature measures per patient. The simultaneous measurements within a given patient should exhibit a high degree of correlation. If .7 is chosen as the lower bound for the correlation between the paired readings (higher correlations would result in an estimate of fewer patient samples thus the choice of .7 is conservative) and if the average standard deviation of the mean of the differences is used as an estimate of the expected variability, then it would be possible to declare measurement equivalence for a difference of ±.19 with a sample size of 40 patients. This assumes the ASTM measurements of instrument variation are representative of the within patient variance. Given the vagaries of patient-to-patient this may be too optimistic an estimate.

A sample size of 40 patients will provide 80% or greater power for testing for equivalence between methods where the mean difference is .1 or less for up to a ten-fold increase in the estimates of the variability of the differences based on the ASTM precision measures and for a mean difference of .15 for up to a five-fold increase in these same measures. If the sample size is 40 then the power for the test of equivalence, where the difference is .15 and the standard deviation is a ten-fold increase, is 35%. Agreement testing: The Bland-Altman test will be used to test the agreement of the two methods. This test will permit a check for significance of bias between the two measures (the average difference between the two measures) as well as a check for significant trending over the range of the measurements. If the bias is significant then the results of the two differ by an overall offset in their measurements. If the trending (slope of the regression line) is significant then the difference between the two measurements changes as their magnitude changes which means the two methods are not in agreement.

The only measurements of instrument precision are those provided by the ASTM assessment. It is reasonable to assume the within patient measurement variation will be greater than the measurements from the controlled testing. A sample size of 40 patients is, from a statistical standpoint, a reasonable number and this sample size provides acceptable power for a mean difference of .1 for a ten-fold increase in the estimate of minimum test variability. This sample size will also provide enough data for the Bland-Altman test for agreement. Because this is a repeated measures test and even one missing measurement will impact the results, the investigators plan to over-sample in order to assure the study has the power to answer the questions posed.