Yes, interevent duration is hours. Here’s a description of this table from http://www.dynsystem.com/NetSTORM/netstormidfsample.htm/ :

The Year table lists annual statistics. The statistics shown are comparable with those in the previous tables with the addition of columns for Events, Non-Events, and Interevent. Events indicates the number of storms exceeding the specified threshold for the year. Non-events are periods of measureable precipitation falling below the threshold. Interevent is the average time (hours) between events. Means of the annual series are listed at the end of the table.

This value is very sensitive to definitions of minimum “storm” depth and distinct event definition (storm event separation time; termed ‘interevent time’ in the existing documentation). The program compiles ‘events’ and ‘non-events’ based on a minimum depth. For instance, if you specify a minimum depth of 0.15”, it discards any storm that fails to meet this criterion. Furthermore, it is very sensitive to how long you specify as an acceptable dry period within a single ‘event.’ For instance, if you specify a storm event separation time of 5 hours and it rains for 3 hours, is dry for 4 hours, and then rains for 6 hours more, the model will tabulate a single event lasting 13 hours. The dry period during the storm is then not considered in determining average interevent duration. (This discussion also demonstrates that I need to change the terminology to indicate the user specifies “storm event separation time” while the model computes “Interevent time ”)

Here’s the existing documentation describing specification of storm event separation time from http://www.dynsystem.com/NetSTORM/help/IDFAnalysis.html/

Interevent time has a major bearing on the listings of "continuous" storm events and plays a minor role in selection of interval events if sub-events are excluded. It is recommended that the interevent time be chosen to obtain a resulting coefficient of variation (standard deviation divided by mean) of 1. NetSTORM displays the coefficient of variation in its results. A trial and error approach should be used to identify the appropriate interevent time. As examples, it has been found to be 12 hours for Dallas, 8 hours in Indianapolis, and 5 hours in Boston. Dry periods longer than the minimum data gap to flag are excluded from the interevent time coefficient of variation calculation. This facilitates exclusion of seasonal dry periods from the computation of this parameter. Thus, with the minimum data gap to flag set to 30 days, an interevent time of 33 hours for Sacramento yields a coefficient of variation of 1.

I suggest that if you need to use the interevent time in an analysis you’re doing, you obtain NetSTORM and tinker with the interevent time and storm minimum total parameters. You may also want to tinker with the ‘minimum gap to flag’ parameter, although this generally has little effect in the eastern US where there is no dry season.