Appendix Table 5.

Common classification trees of flare-related bother¹ derived from 64 participants with interstitial cystitis/bladder pain syndrome and/or chronic prostatitis/chronic pelvic pain syndrome, Washington University School of Medicine and the University of Alabama at Birmingham 2010-2013.

First partition			Second partition (of the first left node)			Second partition (of the first right node)			Percentage of 1,000 trees
Variable	Cut-off point		Variable	Cut-off point		Variable	Cut-off point
	Left node	Right node		Left node	Right node		Left node	Right node
Urgency or frequency	≤6.5	>6.5	Duration	≤One day-long	Multiple days-long	-----	-----	-----	17.3
Urgency or frequency	≤6.5	>6.5	Pelvic pain	≤5.5	>5.5	-----	-----	-----	7.5
Pelvic pain	≤5.5	>5.5	-----	-----	-----	-----	-----	-----	13.2
Pelvic pain	≤5.5	>5.5	-----	-----	-----	Duration	≤Hours-long	≥One day-long	3.2
Pelvic pain	≤6.5	>6.5	Non-urologic pain symptoms	≥1.5	<1.5	-----	-----	-----	1.6
Pelvic pain	≤5.5	>5.5	-----	-----	-----	Duration	Minutes-long	≥Hours-long	1.6
Pelvic pain	≤5.5	>5.5	Duration	≤One day-long	Multiple days-long	-----	-----	-----	1.6
Duration	≤One-day long	Multiple days-long	Urgency or frequency	≤7.5	>7.5	-----	-----	-----	9.3
Duration	Minutes-long	≥Hours-long	-----	-----	-----	Pelvic pain	≤5.5	>5.5	2.9
Duration	≤One-day long	Multiple-days long	Urgency or frequency	≤6.5	>6.5	-----	-----	-----	2.5
Duration	≤Hours-long	≥One day-long	-----	-----	-----	Pelvic pain	≤5.5	>5.5	2.3

¹Defined as “a lot” of general flare-related bother. Each tree can be interpreted as follows: the first partition indicates where the data are first split into two nodes and the second partition indicates where one of these nodes (either the left or right node) is split into two further nodes. Nodes on the right contain mostly flares associated with “a lot” of bother and nodes on the left contain primarily flares associated with not a lot of bother. Each of the partitions is determined by maximizing the association between flare attributes (i.e., intensity of pelvic pain, intensity of urgency or frequency, or duration) and flare-related bother. As an example, we describe the most common tree (first row of the table). The first optimal partition of this tree occurs at values of urgency or frequency >6.5. This means that flares with values of urgency or frequency >6.5 have a higher probability of contributing to a lot of bother than those with values of urgency or frequency ≤6.5. The second and final optimal partition of this tree occurs among flares with values of urgency or frequency ≤6.5 (first left node). Of these flares, those that last for multiple days have a higher probability of contributing to a lot of bother than those that last for one day or less. Thus, considering the tree as a whole, flares with values of urgency or frequency >6.5 and those with values of urgency or frequency ≤6.5 that last for multiple days have a higher probability of contributing to a lot of bother, whereas those with values of urgency or frequency ≤6.5 that last for one day or less have a lower probability of contributing to a lot of bother. This pattern occurred 173 times (17.3%) out of the 1,000 times that the analyses were run with re-sampling.