Skip to main content
. 2022 Feb 2;9(1):e32112. doi: 10.2196/32112

Table 2.

The impact of covariates on the rate of program adoption (Prentice, Williams, and Peterson Gap–Time model estimation resultsa,b).

Variables Program adoption through website Mobile app adoption

HRc,d (95% CI) P value HRd (95% CI) P value
NSESe quintile

First 1.000f N/Ag 1.000f N/A

Second 1.034 (1.015-1.054) .002 0.940 (0.907-0.973) .001

Third 1.029 (1.008-1.051) .02 0.954 (0.918-0.990) .02

Fourth 1.031 (1.009-1.053) .02 0.950 (0.912-0.988) .02

Fifth 1.020 (0.997-1.043) .12 0.948 (0.910-0.987) .02
Age (in years) 1.007 (1.006-1.007) <.001 0.980 (0.979-0.981) <.001
Gender

Female 1.000 N/A 1.000 N/A

Male 1.074 (1.060-1.088) <.001 0.821 (0.797-0.845) <.001
Marketing

No 1.000 N/A 1.000 N/A

Yes 0.378 (0.360-0.396) <.001 17.007 (16.979-17.035) <.001

aInterpreting the estimated hazard ratios, for example, the second neighborhood socioeconomic status (NSES) quintile had an increased likelihood of program adoption via the website by a factor of 1.034 (95% CI 1.015-1.054) as compared with the lowest NSES quintile, keeping other covariates constant (equivalent to a 3.4% increased likelihood of adoption). On the other hand, the likelihood of adoption of the mobile app when comparing the second NSES quintile with the first one shows a decreased likelihood of adoption for the second NSES quintile by a factor of 0.940 (95% CI 0.907-0.973) or 6%.

bObservations=166,932 (the 166,932 observations reflect the 83,466 participants as the model accounts for 2 events per participant); R2=0.255; maximum possible R2=1.000; Wald test (df)=56,343.96 (14); P<.001.

cHR: hazard ratio.

dAn HR of 1.000 was assigned to the reference level for each categorical covariate.

eNSES: neighborhood socioeconomic status.

fFor the HR of 1.000 there is no 95% CI reported, as this is not an estimated HR, but is the default value assigned to the reference level.

gN/A: not applicable (it is the reference level).