Fluvoxamine treatment response prediction in obsessive-compulsive disorder: association rule mining approach

Figures & data

Table 1 Features used in the association rule mining algorithm

No	Feature	No	Feature	No	Feature	No	Feature
1	Female	2	Male	3	Married	4	Early onset (onset <18)
5	Late onset (onset ≥18)	6	No history	7	Family history of mental illness	8	Family history of OCD
9	Aggressive obsession	10	Contamination obsession	11	Sexual obsession	12	Hoarding obsession
13	Religious obsession	14	Symmetry obsession	15	Somatic obsession	16	Cleaning compulsion
17	Checking compulsion	18	Repeating compulsion	19	Counting compulsion	20	Ordering compulsion
21	Obsession severity ≤10	22	Obsession severity >10	23	Compulsion severity ≤10	24	Compulsion severity >10
25	Total severity ≤18	26	Total severity >18	27	Presence of insight	28	No insight
29	Presence of avoidance	30	Lack of avoidance	31	Gaf ≤3	32	Gaf >3
33	Illness duration ≤5 years	34	Illness duration >5 years	35	Responder	36	Non-responder
37	Refractory

Abbreviations: Gaf, global assessment of functioning; OCD, obsessive-compulsive disorder.

Table 2 Pseudo code for Apriori algorithm that is suitable for supervised learning

Function rules=Apriori(T, minsup, minconf, classitem)

% T is the transaction data

% class item is the item that corresponds to class label

F2=Add class item to each element of F1

K=3

Repeat

); {generate candidate k-item sets}

Ck=prune(Ck); {prune candidate item sets}

If t contains c

Count(c)=count(c)+1

end for

end for

Until

//rule generation section

For each candidate item set f

A=f – class item % all items except class item, antecedent part of the rule

B=class item % consequent part of the rule

If confidence(A→B)>minconf

Add A→B to rules

end if

end for

return rules

Table 3 Most important attributes associated with treatment response (support >0.08, lift >1; sorted based on confidence)

Responder		Non-responder		Refractory
Feature	Confidence	Feature	Confidence	Feature	Confidence
Obsession severity ≤10	0.72	Sexual obsession	0.38	Obsession severity >10	0.23
Family history of mental illness	0.72	Illness duration >5 years	0.32	Family history of OCD	0.20
Illness duration ≤5 years	0.68	Being male	0.30	Early onset	0.20
Compulsion severity ≤10	0.67	Compulsion severity >10	0.30	Presence of avoidance	0.18
Married	0.65	Contamination obsession	0.30
Being female	0.64

Abbreviation: OCD, obsessive-compulsive disorder.

Table 4 Best discovered rules for different treatment responses

Class	Rule number	Antecedent	Supp	Conf
Responder	1	Family history of mental illness and Aggressive and Illness duration ≤5	0.16	0.92
	2	Family history of mental illness and Illness duration ≤5	0.19	0.91
	3	Late onset and Family history of mental illness and Illness duration ≤5	0.16	0.89
	4	Female and Married and Late onset and Compulsion severity ≤10	0.16	0.85
	5	Married and Obsession severity ≤10 and Compulsion severity ≤10	0.19	0.82
	6	Married and Hoarding and Obsession severity ≤10	0.16	0.82
Non-responder	7	Contamination and Cleaning and Repeating and Illness duration >5	0.08	0.73
	8	Contamination and Obsession severity >10 and Total severity >18 and Illness duration >5	0.08	0.69
	9	Cleaning and Illness duration >5	0.08	0.69
	10	Cleaning and checking and Obsession severity >10 and Compulsion severity >10	0.07	0.68
	11	Cleaning and checking and Obsession severity >10 and Compulsion severity >10 and Total severity >18	0.07	0.68
Refractory	12	Female and Early onset and Aggressive and Obsession severity >10	0.04	0.83
	13	Female and Married and Early onset and Checking and Gaf ≤3	0.04	0.71
	14	Family history of OCD and Obsession severity >10 and Presence of avoidance and Gaf >3 and Illness duration ≤5	0.04	0.71
	15	Female and Married and Early onset and aggressive and checking and Gaf ≤3	0.03	0.71
	16	Female and Married and Early onset and checking and Gaf ≤3 and Illness duration >5	0.03	0.71

Abbreviations: Conf, confidence; Gaf, global assessment of functioning; OCD, obsessive-compulsive disorder; Supp, support.