Design and statistical analysis reporting among interrupted time series studies in drug utilization research: a cross-sectional survey

Table 3 Basic statistical characteristics of included ITS studies

Characteristics	n	%
Regression model ^a (N = 153)
OLS	47	30.7
ARIMA	24	15.7
GLS	13	8.5
OLS with Newey-West standard errors	12	7.8
Poisson	12	7.8
Mixed effect model	8	5.2
Logistic	4	2.6
Generalized estimating equation	4	2.6
Weighted least square regression	2	1.3
Others ^b	4	2.6
Unclear	23	15.0
Other statistical analysis characteristics
Did the author consider missing data? (N = 153)
Yes	31	20.3
Sensitivity analysis (N = 153)
Yes	49	32.0
Methods for sensitivity analysis (N = 49) ^{c, d}
Transition period	19	38.8
Change measurement of outcomes	10	20.4
Change measurement of study population	11	22.5
Change ITS model setting	9	18.4
Add covariates	2	4.1
Other	11	22.4
Statistical software (N = 153)
SAS	54	35.3
Stata	48	31.4
R	25	16.3
SPSS	5	3.3
Not report	21	13.7
Data availability (N = 153)
Yes	3	2.0
Code availability (N = 153)
Yes	5	3.3

^aThis item refers to the statistical method for the main results in a study
^bOthers included fixed effect model (1 study), negative binomial model (1 study), quasi-poisson model (1 study) and linear probability model (1 study)
^cTransition period: change the interrupted time or time period in the regression model; Change ITS model setting: change the ITS impact model (e.g., from both level and slope change to only level change)
Some studies used more than one method for sensitivity analysis

ISSN: 1471-2288