Thermal Effect of a Woolen Cap in
Low Birthweight Infants during Kangaroo Care

severe hyperthermia (temperature >39 °C) and hypothermia (temperature <35 °C). All SAEs were followed until complete resolution or until the clinician responsible for the care of the recruited patient considered the event to be chronic or the infant to be stable. A monitoring board, including an independent assessor (not involved in the study) from the University of Padua and assessors from each participating hospital, reviewed all the deaths and adverse events. Sample size The lack of information about this topic (temperature during KMC in the specific scenario of head covering) prevented a mathematical estimation of the sample size. Thus, the sample size was arbitrarily defined in order to take into account for the enrollment rate of each hospital. According to the number of admissions in each participating hospital, we decided to enroll a total of 300 subjects (150 in CAP and 150 in NOCAP arms) as follows: 150 at Central Beira Hospital, 90 at St. Luke Wolisso Hospital and 60 at Aber Hospital.[23]
Statistical analysis
The statistician who analyzed the data was blind to treatment allocation. An interim analysis of the first 100 infants was performed as planned in the study protocol.[23] The interim analysis indicated no reasons for stopping for harm (Supplementary File 1), thus the enrollment continued as planned.
The analysis was based on the principle of “intention to treat” and no infants changed treatment arm during the study. Missing data were handled by performing complete case analyses. Among the outcomes, only in-hospital growth included missing data. Continuous data were expressed as mean and standard deviation (SD) or median and interquartile range (IQR), as appropriate.
The effect of the cap on time spent in predefined temperature ranges was assessed using Negative Binomial models, adjusting for hospital and clinically relevant confounders (neonatal temperature at admission to KMC ward, maternal temperature, room temperature, adherence to SSC, birth weight and postnatal age at admission to KMC ward). The models included the logarithm of the number of available readings as an offset (to

take into account for the different lengths of hospital stay of each participant). Negative Binomial model was preferred to Poisson model due to overdispersion.[25]
The effect of the cap on respiratory problems was assessed using a logistic regression model adjusting for hospital and clinically relevant confounders (neonatal and maternal temperatures, room temperature, adherence to SSC, birth weight and postnatal age at KMC admission). The effect of the cap on sepsis was assessed using a logistic regression model adjusting for hospital, adherence to SSC and birth weight. The effect of the cap on in-hospital growth was assessed using a linear regression model adjusting for hospital, adherence to SSC, birth weight and postnatal age at discharge. The low number of deaths prevented any meaningful multivariable analysis of mortality before hospital discharge.
Adherence to SSC was included as confounder in the models because of its variability, previously reported in similar settings.[15] Adherence to SSC was evaluated as the proportion of readings when the neonate was in SSC with the mother over the total number of readings for each neonate.
In addition, a sub-analysis of the effect of the cap on primary outcome was performed in subjects with adherence to SSC ≥ 70% and in those with adherence to SSC < 70%, separately. All regression models included the hospital as covariate, to account for the center-effects.[26] The hospital was included as fixed-effect in the models due to the low number of centers and to the large sample size compared to the number of centers.[26] All tests were 2-sided and a p-value less than 0.05 was considered statistically significant. Statistical analysis was performed using R 3.2.2 software (R Foundation for Statistical Computing, Vienna, Austria).[27]

Results

Patients
From December 2015 to September 2016, 300 infants were enrolled in the study and were randomly assigned to CAP arm (150 infants) or NOCAP arm (150 infants).

Rivista Italiana on line "LA CARE" Volume 11, Numero 1-2, anno 2018