Does frailty impact on length of stay and early readmission following renal transplantation
Paul Hunter, Tara Collidge, John Asher
02/03/2020
Supplementary data for BTS 2020 poster P93
Correlation of variables
Pearson’s r shown for each pair of variables, positive values with blue and negative values with red background.
The transplant length of stay and readmissions within 3 and 6 months are all moderately correlated. Duration of delayed graft function is reasonably correlated with transplant length of stay. The minimum length of stay is 6 days, so this might work better if we modified it to length of stay in excess of six days.
Analysis using a cluster-based model may be appropriate to examine the differences between those likely to be multiply-readmitted versus the almost 50% of transplant patients who do not get any readmissions post-transplant.
Interactions of different variables
The scatterplot matrix shows how different predictor variables interact. The blue circles and curves are for non-diabetic recipients, and the purple triangles and curves for diabetics.
Transplant length of stay
Summary
Tx_LoS represents length of stay in days for the transplant admission.
| Fitness | Tx_LoS.mean | Tx_LoS.sd | Tx_LoS.median | Tx_LoS.iqr | n |
|---|---|---|---|---|---|
| Borderline | 10.6 | 5.5 | 9 | 5 | 35 |
| Fit | 9.9 | 5.8 | 8 | 5 | 106 |
Stratifying by both frailty status and whether delayed graft function (DGF) or immediate function (IF):
| Fitness | DGF_incidence | Tx_LoS.mean | Tx_LoS.sd | Tx_LoS.median | Tx_LoS.iqr | n |
|---|---|---|---|---|---|---|
| Borderline | DGF | 13.8 | 5.9 | 13.5 | 5.8 | 4 |
| Borderline | IF | 10.2 | 5.4 | 9.0 | 4.5 | 31 |
| Fit | DGF | 15.8 | 5.8 | 16.0 | 8.0 | 17 |
| Fit | IF | 8.7 | 5.1 | 7.0 | 3.0 | 89 |
ANCOVA for post-op stay
Anova Table (Type III tests)
Response: Tx_LoS
Sum Sq Df F value Pr(>F)
(Intercept) 1001.6 1 38.6145 6.083e-09 ***
DGF_incidence 587.2 1 22.6367 4.995e-06 ***
Age 74.0 1 2.8544 0.09345 .
DM 171.3 1 6.6030 0.01127 *
Fitness 0.0 1 0.0001 0.99172
Residuals 3475.8 134
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1Effect sizes by \(\eta^2\) shown below:
term etasq
1 DGF_incidence 0.159
2 Age 0.027
3 DM 0.041
4 Fitness 0.000Excess transplant length of stay
Excess transplant stay has been defined as hospital stay in excess of the minimum 6 days.
| Fitness | DGF_incidence | ExcessTxStay.mean | ExcessTxStay.sd | ExcessTxStay.median | ExcessTxStay.iqr | n |
|---|---|---|---|---|---|---|
| Borderline | DGF | 7.8 | 5.9 | 7.5 | 5.8 | 4 |
| Borderline | IF | 4.2 | 5.4 | 3.0 | 4.5 | 31 |
| Fit | DGF | 9.8 | 5.8 | 10.0 | 8.0 | 17 |
| Fit | IF | 2.7 | 5.1 | 1.0 | 3.0 | 89 |
ANCOVA for excess post-op stay
Anova Table (Type III tests)
Response: ExcessTxStay
Sum Sq Df F value Pr(>F)
(Intercept) 347.9 1 13.4115 0.0003589 ***
DGF_incidence 587.2 1 22.6367 4.995e-06 ***
Age 74.0 1 2.8544 0.0934513 .
DM 171.3 1 6.6030 0.0112741 *
Fitness 0.0 1 0.0001 0.9917237
Residuals 3475.8 134
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1\(\eta^2\) effect sizes:
term etasq
1 DGF_incidence 0.159
2 Age 0.027
3 DM 0.041
4 Fitness 0.000Total hospital stay within 3 months
ANCOVA for total length of stay in first 3 months
Anova Table (Type III tests)
Response: Readmissions_LoS_3m
Sum Sq Df F value Pr(>F)
(Intercept) 4.3 1 0.0301 0.862620
DGF_incidence 187.7 1 1.3208 0.252489
Age 1042.6 1 7.3369 0.007631 **
DM 3.9 1 0.0278 0.867900
Fitness 1.2 1 0.0086 0.926172
Residuals 19184.3 135
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1The \(\eta^2\) effect size shows that frailty score has minimal effect on length of stay in first 3 months. While age was a statistically significant predictor, the effect size was relatively small.
term etasq
1 DGF_incidence 0.011
2 Age 0.053
3 DM 0.000
4 Fitness 0.000Total hospital stay within 6 months
ANCOVA for hospital stay in first 6 months
Anova Table (Type III tests)
Response: Readmissions_LoS_6m
Sum Sq Df F value Pr(>F)
(Intercept) 99.1 1 0.5473 0.4607193
DGF_incidence 195.6 1 1.0799 0.3005734
Age 2242.9 1 12.3847 0.0005905 ***
DM 5.2 1 0.0289 0.8653438
Fitness 13.9 1 0.0769 0.7819753
Residuals 24449.1 135
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1By this stage, delayed graft function has almost no impact, but age is a statistically significant predictor, albeit with a small \(\eta^2\) effect size:
term etasq
1 DGF_incidence 0.009
2 Age 0.085
3 DM 0.000
4 Fitness 0.001MANCOVA for post-transplant stay and readmissions within 3 and 6 months
Response Tx_LoS :
Df Sum Sq Mean Sq F value Pr(>F)
Age 1 129.8 129.77 5.0097 0.026880 *
DM 1 292.8 292.80 11.3032 0.001013 **
Fitness 1 6.9 6.92 0.2672 0.606100
DGF_incidence 1 588.0 588.02 22.6999 4.92e-06 ***
Age:Fitness 1 55.6 55.59 2.1461 0.145314
Residuals 132 3419.3 25.90
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
Response ExcessTxStay :
Df Sum Sq Mean Sq F value Pr(>F)
Age 1 129.8 129.77 5.0097 0.026880 *
DM 1 292.8 292.80 11.3032 0.001013 **
Fitness 1 6.9 6.92 0.2672 0.606100
DGF_incidence 1 588.0 588.02 22.6999 4.92e-06 ***
Age:Fitness 1 55.6 55.59 2.1461 0.145314
Residuals 132 3419.3 25.90
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
Response Readmissions_LoS_3m :
Df Sum Sq Mean Sq F value Pr(>F)
Age 1 1134.1 1134.06 7.8564 0.005828 **
DM 1 9.4 9.36 0.0648 0.799388
Fitness 1 9.8 9.78 0.0678 0.794993
DGF_incidence 1 176.6 176.57 1.2232 0.270741
Age:Fitness 1 36.1 36.06 0.2498 0.618019
Residuals 132 19054.1 144.35
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
Response Readmissions_LoS_6m :
Df Sum Sq Mean Sq F value Pr(>F)
Age 1 2385.3 2385.30 13.0043 0.0004395 ***
DM 1 6.2 6.21 0.0339 0.8542617
Fitness 1 35.5 35.54 0.1938 0.6605085
DGF_incidence 1 181.4 181.37 0.9888 0.3218574
Age:Fitness 1 69.8 69.81 0.3806 0.5383413
Residuals 132 24211.9 183.42
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1Age, diabetes and delayed graft function are all statistically significant predictors of duration of post-op inpatient stay and inpatient stay in excess of 6 days, but only age is a predictor of length of stay for readmissions. Neither fitness nor the interaction of age and fitness were predictors.