Table 1. Factors Predictive of Mortality Among Human Immunodeficiency Virus-negative Children 1-59 Months of Age Presenting to Hospital With Severe or Very Severe Pneumonia: Multivariable Analyses.
| Characteristic | Died Within 7 d of Discharge | Unadjusted | Adjusted | ||||||
|---|---|---|---|---|---|---|---|---|---|
| No | Yes | ORa | (95% CI) | aOR | (95% CI) | LRT P Valuea | |||
| Allb | 1682 | (93.3) | 120 | (6.7) | ... | ... | ... | ... | |
| Site | |||||||||
| Kenya | 372 | (96.6) | 13 | (3.4) | 1.06 | (.48-2.35) | 1.13 | (.41-3.11) | |
| The Gambia | 235 | (95.5) | 11 | (4.5) | 1.42 | (.61-3.26) | 2.13 | (.73-6.27) | |
| Mali | 213 | (86.6) | 33 | (13.4) | 4.69 | (2.37-9.27) | 2.20 | (.84-5.75) | |
| Zambia | 125 | (74.0) | 44 | (26.0) | 10.6 | (5.45-20.8) | 12.3 | (5.13-29.7) | |
| South Africa | 363 | (96.8) | 12 | (3.2) | 1 | (ref) | 1 | (ref) | |
| Thailand | 115 | (97.5) | 3 | (2.5) | 0.80 | (.22-2.85) | 0.77 | (.09-6.69) | |
| Bangladesh | 259 | (98.5) | 4 | (1.5) | 0.47 | (.15-1.46) | 1.00 | (.25-3.99) | |
| Age, mo | |||||||||
| 1-11 | 947 | (91.9) | 84 | (8.1) | 1.37 | (.89-2.10) | 2.20 | (1.28-3.78) | .0031 |
| 12-59 | 735 | (95.3) | 36 | (4.7) | 1 | (ref) | 1 | (ref) | |
| Sex | |||||||||
| Male | 974 | (95.0) | 51 | (5.0) | 1 | (ref) | 1 | (ref) | .0038 |
| Female | 708 | (91.1) | 69 | (8.9) | 1.76 | (1.19-2.61) | 1.99 | (1.24-3.20) | |
| Unresponsiveness and/or deep breathingc | |||||||||
| Neither | 1149 | (95.2) | 58 | (4.8) | 1 | (ref) | 1 | (ref) | <.0001 |
| Deep breathing, but alert | 366 | (95.8) | 16 | (4.2) | 1.46 | (.78-2.73) | 1.18 | (.58-2.39) | |
| Unresponsive but no deep breathing | 123 | (83.7) | 24 | (16.3) | 4.61 | (2.60-8.19) | 3.12 | (1.51-6.45) | |
| Unresponsive and deep breathing | 36 | (62.1) | 22 | (37.9) | 19.4 | (9.74-38.8) | 14.6 | (6.53-32.8) | |
| Cough (observed) | |||||||||
| No | 498 | (88) | 68 | (12) | 1 | (ref) | 1 | (ref) | .0032 |
| Yes | 1175 | (95.8) | 52 | (4.2) | 0.43 | (.28-.65) | 0.48 | (.29-.78) | |
| Grunting (observed) | |||||||||
| No | 1414 | (95.8) | 62 | (4.2) | 1 | (ref) | 1 | (ref) | .0026 |
| Yes | 257 | (81.6) | 58 | (18.4) | 2.77 | (1.67-4.58) | 2.48 | (1.37-4.48) | |
| Hypoxemiad | |||||||||
| No | 1125 | (95.7) | 51 | (4.3) | 1 | (ref) | 1 | (ref) | .0002 |
| Yes | 554 | (88.9) | 69 | (11.1) | 3.18 | (2.08-4.86) | 2.55 | (1.54-4.22) | |
| Maximum duration of illness,e d | |||||||||
| 0-2 | 637 | (96.5) | 23 | (3.5) | 1 | (ref) | 1 | (ref) | .0018 |
| 3-5 | 740 | (93.4) | 52 | (6.6) | 1.87 | (1.11-3.15) | 2.15 | (1.16-3.99) | |
| >5 | 291 | (87.4) | 42 | (12.6) | 3.21 | (1.84-5.63) | 3.28 | (1.66-6.46) | |
| Weight-for-height z-score | |||||||||
| Very low (< -3) | 160 | (82.5) | 34 | (175) | 4.55 | (2.75-755) | 3.57 | (2.03-6.31) | <.0001 |
| Low (≥ -3 to < -2) | 223 | (91) | 22 | (9) | 2.54 | (1.47-4.41) | 2.45 | (1.32-4.52) | |
| Normal-high (≥ -2) | 1258 | (96) | 52 | (4) | 1 | (ref) | 1 | (ref) | |
Data are presented as no. (%) unless otherwise indicated. Abbreviations: aOR, adjusted odds ratio; CI, confidence interval; d, days; LRT, likelihood ratio test; OR, odds ratio; ref, reference category.
Children presenting to hospital with cough or difficulty breathing (observed or history of) and observed lower chest wall indrawing or at least 1 danger sign from the World Health Organization Pocketbook of Hospital Care for children were enrolled into the Pneumonia Etiology Research for Child Health (PERCH) study. The full set of univariable analyses, including information on missing data, is included in Supplementary Table 3; factors displayed in this table are only those that remained in the final multivariable model. All analyses controlled for site using “site” as a forced, indicator variable. P values were obtained from logistic regression LRT During backwards regression modeling, covariates were removed from the model if they did not significantly improve the fit of the model to the data (P > .05). When variables associated with the outcome in univariable analyses were added back into the model, they did not significantly increase the fit of the model to the data. When the model was run using random effects to control for clustering rather than logistic regression estimating fixed effects for each site, the coefficients were identical to those above and the allocated scores were the same.
A total of 1719 of 1802 observations were used in the final model (95%); observations with missing data were significantly associated with higher mortality (x 2 P < .05), and their omission from the final model reduced the overall mortality from 6.7% to 6.2%.
LRT P value for interaction = .018.
Hypoxemia was defined as oxygen saturation <92% in all sites except Zambia and South Africa (the 2 sites situated at altitude >1000 m), where it was defined as <90%.
Maximum reported duration of illness with fever, cough, difficulty breathing, or wheeze.