Table 4.
Characteristics for the clusters using non-negative matrix factorisation technique among asthmatics (NLP-PAC+/NLP-API+, NLP-PAC+ only and NLP-API+ only group) (the three clusters were depicted in the heatmap analysis in figure 2)
| Cluster A (n=655) | Cluster B (n=843) | Cluster C (n=1175) | Total (n=2673) | P value | |
| Group defined by NLP, n (%) | <0.001* | ||||
| Group 1 (NLP-PAC+/NLP-API+) | 536 (82) | 430 (51) | 648 (55) | 1614 (60) | |
| Group 2 (NLP-PAC+ only) | 110(17) | 369 (44) | 475 (40) | 954 (36) | |
| Group 3 (NLP-API+ only) | 9 (1) | 44 (5) | 52 (4) | 105 (4) | |
| Male, n (%) | 369 (56) | 442 (52) | 727 (62) | 1538 (58) | <0.001* |
| Ethnicity, n (%) | <0.001* | ||||
| White | 575 (88) | 765 (91) | 805 (69) | 2145 (80) | |
| HOUSES at birth in lowest quartile, n (%) | 114 (17) | 112 (13) | 331 (28) | 557 (21) | <0.001* |
| Overweight†, n (%) | 108 (17) | 176 (21) | 196 (17) | 480 (18) | 0.021* |
| Unknown, n (%) | 2 (0) | 13(2) | 9 (1) | 24 (1) | |
| Maternal smoking history during pregnancy, n (%) | 42 (6) | 63 (8) | 126 (11) | 231 (9) | <0.001* |
| Unknown, n (%) | 44 (7) | 104 (12) | 162 (14) | 310 (12) | |
| Infrequent well-child visit per year in quartile in the lowest quartile, n (%) | 101 (15) | 332 (39) | 235 (20) | 668 (25.0) | <0.001* |
| Family history of asthma, n (%) | 242 (37) | 244 (29) | 271 (23) | 757 (28) | <0.001* |
| Early onset asthma‡, n (%) | 475 (73) | 532 (63) | 1031 (88) | 2038 (76) | <0.001* |
| Eczema, n (%) | 329 (50) | 254 (30) | 400 (34) | 983 (37) | <0.001* |
| Allergic rhinitis, n (%) | 389 (59) | 22 7(27) | 210 (18) | 826 (31) | <0.001* |
| Eosinophilia, n (%)§ | 271 (41) | 147 (17) | 300 (26) | 718 (27) | <0.001* |
| Unknown, n (%) | 153 (23) | 329 (39) | 518 (44) | 1000 (37) | |
| Total IgE (kU/L)>300, n (%) | 41 (6) | 7 (1) | 11 (1) | 59 (2) | 0.32* |
| Unknown, n (%) | 520 (79) | 809 (96) | 1124 (96) | 2453 (92) | |
| Elevated IgE to any aeroallergen, n (%)¶ | 48 (7) | 11 (1) | 14 (1) | 73 (3) | 0.68* |
| Unknown, n (%) | 535 (82) | 813 (96) | 1132 (96) | 2480 (93) | |
| FEV1/FVC<0.85, n (%) | 238 (36) | 136 (16) | 67 (6) | 441 (17) | 0.65* |
| Unknown, n (%) | 237 (36) | 590 (70) | 1050 (89) | 1877 (70) | |
| Acute exacerbation of asthma, n (%)** | 262 (40) | 77 (9) | 96 (8) | 435 (16) | <0.001* |
| HEDIS††-defined persistent asthma, n (%) | 306 (47) | 123 (15) | 107 (9) | 536 (20) | <0.001* |
| Pneumonia, n (%) | 301 (46) | 179 (21) | 327 (28) | 807 (30) | <0.001* |
| PE tube, n (%) | 106 (16) | 57 (7) | 122 (10) | 285 (11) | <0.001* |
| Pertussis, n (%) | 23 (4) | 19 (2) | 20 (2) | 62 (2) | 0.047* |
| Zoster, n (%) | 17 (3) | 23 (3) | 14 (1) | 54 (2) | 0.026* |
| Appendicitis, n (%) | 11 (2) | 23 (3) | 13 (1) | 47 (2) | 0.023* |
| Coeliac disease, n (%) | 9 (1) | 6 (1) | 11 (1) | 26 (1) | 0.42* |
| Frequent viral infection per year in group in the highest quartile, n (%) | 403 (62) | 13 (2) | 253 (22) | 669 (25) | <0.001* |
| Frequent strep infection per year in the highest quartile, n (%) | 323 (49) | 9 (1) | 102 (9) | 434 (16) | <0.001* |
*Pearson’s χ2 test.
†Overweight at asthma index date (for children age 2 years or more, BMI for age at or above 85% used (https://www.cdc.gov/obesity/childhood/defining.html)).
‡Early onset asthma defined by age at either criteria met date <6 years old.
§Eosinophilia defined by >300/µL (PAC) or ≥4% (API).
¶Elevated IgE defined by >0.35 kU/L to any aeroallergen among alternaria tenuis, cat epithelium, dog dander, house dust mite/D.F., house dust mite/D.P., elm, oak, short ragweed, timothy grass.
**Acute exacerbation of asthma defined by any of ER visit, hospitalisation or systemic corticosteroid use for asthma during follow-up period.
††Healthcare Effectiveness Data and Information Set.
‡‡PE tube, pressure equalising tube as a surrogate marker for frequent ear infection.
§§Strep infection, Streptococcus pyogenes upper respiratory infection.
¶¶Season: spring (March–May), summer (June–August), autumn (September–November) and winter (December, January and February).
BMI, body mass index; HOUSES, HOUsing-based Index of SocioEconomic Status; NLP, natural language processing; NLP-API, NLP algorithms for Asthma Predictive Index; NLP-CAP, NLP algorithms for Predetermined Asthma Criteria.