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

Table 2 Design characteristics (rationale for ITS, data handling and model structure) of included studies

Characteristics	n or median	% or IQR
Rationale for ITS
Did the author give the reason/rationale for using ITS study design? (N = 153)
Yes	43	28.1
Clearly segment time? (N = 153)
Yes	153	100.0
Use of control group (N = 153)
Yes	19	12.4
Type of control (N = 19)^a
Characteristic	11	57.9
Location	6	31.6
Historical	1	5.3
Outcome	1	5.3
Data collection and aggregation
Data collection (N = 153)
Retrospective	141	92.2
Prospective	12	7.8
For studies used prospective data, if they have pre-specified protocol? (N = 12)
Yes	2	16.7
Raw data (N = 153)
Contained individual-level data	127	83.0
Only contained aggregated-level data	26	17.0
Analysis unit ^b (N = 153)
Aggregated unit	149	97.4
Individual unit	4	2.6
Time intervals used for ITS with aggregated unit (N = 149)
Month	110	73.8
Quarter	17	11.4
Week	7	4.7
Year	6	4.0
Day	4	2.7
Six-month	2	1.3
Two-week	2	1.3
Two-month	1	0.7
Number of time points (N = 149)^c	48	30, 72
Did the author give the rationale for the number of time points? (N = 149)
Yes	31	20.8
The ITS model structure
Type of ITS impact model (N = 153)
Level and slope change	139	90.8
Level change only	6	3.9
Slope change only	6	3.9
Unclear	2	1.3
Did the author give the rationale for using this model? (N = 153)
Yes	21	13.7
Did the author allow for a delay impact? (N = 153)
Yes	42	27.5
Methods of dealing with delay (N = 42)^{d, e}
Sensitivity	19	45.2
Segment	11	26.2
Excluded	10	23.8
Delay	6	14.3
Did the author give the rationale for setting the transition period? (N = 42)
Yes	23	54.8

^aLocation: use a different area as control; Outcome: use an outcome not affected by the intervention as control; Characteristic: use a group not targeted by an intervention as control; Historical: compare a previous group to a current group
^bAggregated data refer to summary statistics (e.g., mean, percentage, median) calculated across individual data
^cWe reported the median and interquartile range (25% and 75%)
^dDelay: where the delay was acknowledged and included in pre- or post-interruption segment; Excluded: where a separate segment was used for the delay time period, but this was excluded from the analysis; Segment: where a separate segment was used for the delay time period, and this was included in analysis; Sensitivity: where the delay was modelled as part of a sensitivity analysis, but ignored in main analysis)
^eSome studies used more than one method to deal with delay

ISSN: 1471-2288