Table 3

Growth curve model of pulse oximetry adoption and orders for oxygen use at admission

PredictorsPulse oximetry done?Oxygen prescribed?
OR95% CIP valueOR95% CIP value
(Intercept)*0.030.01 to 0.17<0.0010.010.01 to 0.02<0.001
Hospital factors
 Malaria endemicity zone: high (ref: low)1.620.1 to 25.040.7291.550.9 to 2.670.112
 Time (months in CIN)1.111.05 to 1.18<0.0011.021 to 1.030.045
Patient-level factors
Ref: patient in any study (No)†
 Patient in any study (prestudy period)‡2.792.58 to 3.02<0.0011.020.9 to 1.150.800
 Patient in any study (Yes)1.461.35 to 1.58<0.0011.020.91 to 1.150.715
 Referral: Yes (ref: No)1.051.01 to 1.090.0251.191.12 to 1.26<0.001
 Readmission: Yes (ref: No)1.010.97 to 1.050.7321.071 to 1.150.052
 Is a weekend admission: Yes (ref: No)0.980.95 to 1.010.1441.071.02 to 1.130.009
 PAR used: Yes (ref: No)1.991.79 to 2.21<0.0011.231.02 to 1.480.031
 PAR used * time (months in CIN)1.011 to 1.01<0.00111 to 1.010.562
Ref: age (>59 months)
 Age (<1 month)0.460.41 to 0.51<0.0011.961.62 to 2.35<0.001
 Age (1–11 months)1.081.03 to 1.13<0.0011.431.32 to 1.54<0.001
 Age (12–59 months)1.091.05 to 1.13<0.0011.081 to 1.160.054
 Female: Yes (ref: No)10.97 to 1.020.7231.020.97 to 1.060.424
 Fever: Yes (Ref: No)1.181.14 to 1.23<0.0010.780.73 to 0.83<0.001
 Cough: Yes (Ref: No)1.041.01 to 1.080.0081.361.29 to 1.44<0.001
 Difficulty breathing: Yes (ref: No)1.061.02 to 1.10.0022.792.64 to 2.94<0.001
 Stridor: Yes (ref: No)1.050.95 to 1.160.3251.141.02 to 1.270.017
 Cyanosis: Yes (ref: No)0.770.65 to 0.920.0041.51.27 to 1.77<0.001
 Indrawing: Yes (ref: No)1.251.2 to 1.31<0.0012.362.23 to 2.49<0.001
 Grunting: Yes (ref: No)0.880.84 to 0.93<0.0011.91.8 to 2<0.001
 Acidotic breathing: Yes (ref: No)0.890.8 to 0.980.0181.341.22 to 1.47<0.001
 Wheezing: Yes (ref: No)1.050.98 to 1.120.1671.131.06 to 1.210.001
 Crackles: Yes (ref: No)1.061.01 to 1.110.0101.251.19 to 1.32<0.001
 Tachypnoea: Yes (ref: No)0.960.93 to 0.990.0221.591.51 to 1.68<0.001
 Alert (AVPU<A): No (ref: Yes)0.960.9 to 1.030.2381.951.81 to 2.1<0.001
 Inability to drink: Yes (ref: No)1.041 to 1.080.0430.80.76 to 0.85<0.001
Ref: pallor (none)
 Pallor (mild/moderate)0.930.89 to 0.970.0021.11.02 to 1.170.008
 Pallor (severe)0.940.89 to 0.990.0241.31.19 to 1.42<0.001
Ref: WAZ (normal)
 WAZ (low)0.970.92 to 1.010.1350.980.91 to 1.050.551
 WAZ (very low)0.940.9 to 0.990.0200.930.87 to 10.059
 Slow capillary refill: Yes (ref: No)0.880.78 to 10.0510.970.83 to 1.130.693
Ref: hypoxaemia (unknown)
 Hypoxaemia (No)0.590.56 to 0.62<0.001
 Hypoxaemia (Yes)1.921.77 to 2.08<0.001
  • *Represents the average odds that a patient will have pulse oximetry is done (or oxygen therapy is prescribed) at admission in this sample when all the predictors are set to their reference levels (eg, ‘No’) at the first month of being a member of the Clinical Information Network (CIN).

  • †'Patient is in any study’ point estimates and CIs for the ‘pulse oximetry done?’ outcome do not overlap with ones from complete case analysis (online supplemental table 4). From the complete case sensitivity analysis, there is a material difference in the estimated association of involvement in any research study variable with the pulse oximetry use at admission outcome (table 3, online supplemental table 4); while the direction of association is the same (positive), the CIs do not overlap. With the age variable, the direction of the association is inverted between the complete case and multiple imputation analyses, and children aged <1 month admitted to the paediatric ward are completely missing in the complete case analysis. These material difference might be attributable to the fact that complete case analysis and our multiple imputation approach make different assumptions of missingness mechanism in the data (ie, missing completely at random mechanism—which is potentially more biased, and a missing at random mechanism,respectively).

  • ‡'Patient is in any study (Pre-study period)’ reflects the time leading up to any research study starting when no study was being conducted, that is, February 2014–November 2018.

  • AVPU, Alert, Verbal, Pain, Unresponsive; WAZ, Weight-for-age Z-score.