🆕 Aim 2 Stratified PMA vs FOF
Executive Summary¶
Aim 2: Examine the relationship between PMA at first oral feeding and time to full oral feeding, stratified by before vs after 35 weeks PMA.
Sample: 228 infants (166 <35w, 62 ≥35w)
Key Findings: - Unadjusted interaction: p = 0.0320 (significant) - Adjusted interaction: p = 0.0519 (not significant)
Overall Relationship (Whole Cohort)¶
Whole-cohort relationship between pma_at_first_oral_feeding and time_to_fof (n=228).
Correlation¶
- Pearson r = 0.006 (p = 0.9250)
Linear Model¶
- Slope: 0.056 (SE = 0.591, p = 0.9250)
- R²: 0.0000
Quadratic Model¶
- R²: 0.0208
- Linear term: -36.435 (p = 0.0304)
- Quadratic term: 0.5182 (p = 0.0301)
- Minimum at PMA = 35.2 weeks
Stratified Linear Regression (Unadjusted)¶
Separate linear regressions of pma_at_first_oral_feeding on time_to_fof for infants who started feeding before vs at/after 35 weeks PMA.
Group-Specific Slopes¶
| Group | n | Slope | SE | p-value | 95% CI | R² |
|---|---|---|---|---|---|---|
| <35w | 166 | -2.353 | 1.222 | 0.0559 | (-4.765, 0.060) | 0.0221 |
| ≥35w | 62 | 1.606 | 1.48 | 0.2821 | (-1.354, 4.567) | 0.0193 |
Formal Interaction Test¶
Model: time_to_fof ~ pma_at_first_oral_feeding * group_35w
- Interaction coefficient: 3.959 (SE = 1.835)
- Interaction p-value: 0.0320
- Full model R²: 0.0221
- N: 228
Model Diagnostics¶
- Condition number: 3022.7
- Multicollinearity warning: Yes
- Normality (Jarque-Bera p): < 0.0001
- Heteroscedasticity (Breusch-Pagan p): 0.2584
Stratified Linear Regression (Adjusted)¶
Same stratified analysis adjusted for: gestational_age_weeks, birth_weight_grams, mechanical_ventilation, o2_device_at_36_weeks.
Adjusted Group-Specific Slopes¶
| Group | n | Adj. Slope | SE | p-value | 95% CI | R² |
|---|---|---|---|---|---|---|
| <35w | 166 | -3.275 | 1.232 | 0.0086 | (-5.707, -0.842) | 0.1905 |
| ≥35w | 62 | -0.2 | 1.596 | 0.9008 | (-3.396, 2.996) | 0.185 |
Adjusted Interaction Test¶
- Adjusted interaction coefficient: 3.509 (SE = 1.795)
- Adjusted interaction p-value: 0.0519
- Adjusted model R²: 0.1508
Adjusted Model Diagnostics¶
- Condition number: 109511.3
- Multicollinearity warning: Yes
Stratified Quadratic Regression¶
Quadratic models within each group to test whether optimal PMA differs.
<35w¶
- n = 166
- R²: 0.0519
- Linear term: -175.875 (p = 0.0230)
- Quadratic term: 2.5898 (p = 0.0249)
- Minimum at PMA = 34.0 weeks
≥35w¶
- n = 62
- R²: 0.0309
- Linear term: 71.832 (p = 0.3935)
- Quadratic term: -0.9449 (p = 0.4041)
- No minimum within data range
Visualization¶
Scatter plot with separate regression lines and 95% confidence bands per group.
Summary¶
Does the relationship differ before vs after 35 weeks PMA?¶
Unadjusted slopes: - <35w: -2.353 (p = 0.0559) - ≥35w: 1.606 (p = 0.2821) - Interaction p = 0.0320
Adjusted slopes: - <35w: -3.275 (p = 0.0086) - ≥35w: -0.200 (p = 0.9008) - Adjusted interaction p = 0.0519
Reconciliation with Prior Reports¶
- Reports 10/11: Used pma_at_full_feed as outcome (not time_to_fof). Found significant slope difference (p=0.032 unadjusted, p=0.052 adjusted).
- Report 15: Used categorical early/late predictor. Found no difference in absolute time_to_fof (p=0.638).
- This report: Uses continuous pma_at_first_oral_feeding predicting time_to_fof with formal interaction test, directly answering Aim 2.