Interpreting patient-Specific risk prediction using contextual decomposition of BiLSTMs: application to children with asthma

Table 4 Top scoring patterns of length 2 visit, produced by the contextual decomposition of LSTM and BiLSTM models on the asthma data

	LSTM		BiLSTM
	ICD Codes	Frequency%	ICD Codes	Frequency%
1	[493.9],[493.9] [Asthma Unspecified],[Asthma Unspecified]	13%	[493.9], [493.9] [Asthma Unspecified],[Asthma Unspecified]	11%
2	[493.9,786.0],[493.9][Asthma Unspecified, Dyspnea and Respiratory Ab-normalities], [Asthma Unspecified]	2%	[493.9,786.0],[493.9][Asthma Unspecified, Dyspnea and Respiratory Ab-normalities], [Asthma Unspecified]	2%
3	[493.9],[493.9,786.0] [Asthma Unspecified], [Asthma Unspecified, Dysp-nea and Respiratory Abnormalities]	2%	[493.9],[493.9,786.0][Asthma Unspecified], [Asthma Unspecified, Dysp-nea and Respiratory Abnormalities]	2%
4	[493.9], [V20.2] [Asthma Unspecified], [Routine Infant or Child Health Check]	2%	[493.9], [V20.2][Asthma Unspecified], [Routine Infant or Child Health Check]	2%
5	[493.9,786.2], [493.9] [Asthma Unspecified, Cough], [Asthma Unspecified]	2%	[493.9,786.2], [493.9][Asthma Unspecified, Cough], [Asthma Unspecified]	1%

ISSN: 1472-6947