Table. Characteristics of the Analysis and the Outlier (Excluded) Sets as Generated by LIMIT and the Overall Samplea.
| Characteristic | Patient group | ||
|---|---|---|---|
| Analysis | Outlier | Overall | |
| Temperature, mean (95% CI), °Cb | 36.64 (35.95-37.33) | 36.71 (35.87-37.55) | 36.66 (35.92-37.41) |
| Among men only | 36.59 (35.90-37.28) | 36.67 (35.81-37.53) | 36.62 (35.86-37.38) |
| Among women only | 36.67 (36.00-37.34) | 36.74 (35.92-37.56) | 36.69 (35.97-37.42) |
| No. of encounters | 396 196 | 222 110 | 618 306 |
| No. of patientsc | 126 705 | 86 744 | 151 977 |
| Sex, No. (%) | |||
| Men | 54 035 (42.65) | 34 569 (39.85) | 65 117 (48.85) |
| Women | 72 670 (57.35) | 52 175 (60.15) | 86 860 (57.15) |
| Visits per person, median (IQR) [range]d | 2.0 (1-4) [1-87] | 2.0 (1-3) [1-54] | 2.0 (1-5) [1-111] |
| Age, mean (SD), yb | 52.7 (15.9) | 54.2 (15.6) | 53.2 (15.8) |
| Height, mean (SD), mb,e | 1.68 (0.10) | 1.67 (0.10) | 1.68 (0.01) |
| Weight, median (IQR) [range], kgb,d,e | 75.3 (63.1-88.9) [30.4-181.1] | 77.4 (64.8-92.3) [30.4-181.4] | 76.1 (63.6-90.3) [30.4-181.4] |
| Encounters by department, No. (%) | |||
| Family medicine | 223 194 (56.33) | 115 790 (52.13) | 338 984 (54.82) |
| Internal medicine | 173 002 (43.67) | 106 320 (47.87) | 279 322 (45.18) |
| Time of day, median (IQR) [range], 24-h clockb,d | 11 (9-14) [0-20] | 11 (10-14) [0-23] | 11 (10-14) [0-23] |
Abbreviation: LIMIT, Laboratory Information Mining for Individualized Thresholds algorithm.
Indicates raw data (without adjustment). The distributions of all variables were statistically significantly different between analysis and outlier sets (P < .001 for all).
Indicates per encounter.
The same individual can have encounters in both the analysis and the outlier sets.
Indicated for those variables not demonstrating a normal distribution.
A total of 16 757 (4.2%) encounters had neither weight nor height values; these were approximated by the nearest other encounter if available (eAppendix in Supplement 1). Note that among adults, changes in height between encounters are likely minimal whereas weight can vary significantly. A total of 130 199 (32.9%) and 3412 (0.9%) values were missing for height only or weight only, respectively; these were calculated based on body mass index.