LA METODOLOGÍA A APLICAR

Patient 6 (Figure 3.7) did not survive their stay in the hospital. Over the course of 399 hours, patient 6 received fluids by IV, oral intake, and continual infusion. kc,vol indicates a potentially low total fluid volume throughout their stay. From t = 140 h to t = 325 h, SCr increased by 1.8 mg L-1 day-1. During this time period, there was minimal change in urine production rate. This complements the model’s assessment of low fluid levels. According to subplot ‘b’, the rate of continual infusion of fluids for patient 6 was 0.075 L h-1 which is

less then the rate of fluid intake required for the model to maintain equilibrium. Fluid levels slowly decrease throughout the patient’s 16 days. Subplot ‘e’ shows no loss of fluid by way of Ve(t). Studying all ’b’ subplots, the collective rate of fluid consumption for patient 6 is significantly less then any of the other 10 patients. ∆SCr for this patient was 12 mg L-1_. Subplot ‘g’ shows a gradual decrease in kidney function over time - from 55% to roughly 35%. SCr and UOT RMSE values indicate a good fit and is comparable to the results of the surviving patients. Minimal oscillatory motion is seen with GFRs(t) which further indicates a good fit.

Figure 3.7: Patient 6 (fatality) data and fitted results. Time in hospital = 399 h.

Similar to patient 6, patient 7 (Figure 3.8) was in the hospital for 398 h.This patient received fluid via six methods: IV, oral intake, continual infusion, blood products / colloids, feeding tube, and enteral intake. kc,vol predicts this patient to have fluid levels greater than

100% upon their arrival to the hospital. Patient 6 continually lost fluid via the interstitial compartment which caused a decrease in fluid levels during their time in the hospital. At time points 180 h, 210 h, 260 h, and 300 h, significant fluid was lost through Ve(t). To initially counter these losses, IV was administered at an increased rate. At t = 260 h, significant fluid loss in Ve(t) was countered by blood products / colloids administered at a rate of 1.15 L h-1_{. The total change in SCr is 27 mg L}-1_{. The data shows a gradual increase in SCr} over time. The model does not fit SCr values well for the first 100 h which is problematic. This may be due to GFRs(t) being zero at t = 0 h and creating a significant accumulation of creatinine in a short amount of time. The amount of oscillation present in GFRs(t) over time is not ideal, however, its ability to fit UOT over time was good. Likewise, SCr values were fit well at t > 100 h. SCr and UOT RMSE values for patient 7 are 5.35 mg L-1 and 2.08 L, respectively. The RMSE for SCr was the highest value recorded for all 10 patients and is attributed to the performance of the model over the first four days. UOT RMSE was also exceptionally high, but relatively similar to patients 8 and 9. Underlying conditions for patient 7 not captured by the model may have influenced the results seen here. Increasing the limitations of GFRs(t) would also improve the fit performance.

Figure 3.8: Patient 7 (fatality) data and fitted results. Time in hospital = 398 h.

Patient 8 was in the hospital for 470 hours (19.5 days). The results of the fit for this patient are given in Figure 3.9. Seven methods of fluid input were utilized: IV, oral intake, continual infusion, blood products / colloids, operation intake, feeding tube, and enteral intake. Enteral intake was steady over the first 400 hours of their stay and then decreased. A feeding tube was utilized after their first 130 hours in the hospital. interstitial fluid was continually lost during their time. Highly notable losses of fluid occurred at t = 180 h, 350 h, and 400 h. Its impact on the interstitial compartment and Vtot(t) are apparent. The urine output rate over the first 10 days in the hospital was very low and the model was not capable of capturing it. An underlying condition in this patient is not accounted for by the

model. The model focused on minimizing the error in SCr as a result of the poor fit to UOT. The total change in SCr was 52 mg L-1_{. For patient 8, SCr and UOT RMSE values are 1.16} mg L-1 _{and 1.84 L. This fit could be improved upon in the future.}

Figure 3.9: Patient 8 (fatality) data and fitted results. Time in hospital = 470 h.

Firgure 3.10 gives the fitted results for patient 9 during their 378 hours in the hospital. Five different methods of fluid delivery were used for this individual: IV, oral intake, contin- ual infusion, feeding tube, and enteral intake. After t = 240 h, the only form of fluid delivery was oral intake. Upon admittance, significant rates of fluid delivery were administered to the patient via IV. The interstitial compartment was losing significant amounts of fluid at the same time. kc,vol indicated close to 100% fluid hydration at t = 0 h. The significant losses of fluid via Ve(t) countered that hydration level. Upon hospital admission, SCr continually increased and GFRs(t) ranged from 0.0 to 0.1 over the first 3 days. GFRs(t) slowly improved

over time. SCr ranged from 17 mg L-1_{to 92 mg L}-1_{. The effectiveness of the model to capture} the data for patient 9 was generally good. SCr and UOT RMSE values were 5.67 mg L-1 and 1.84, respectively. But, the fairly high RMSE for SCr is due to two data points missed. The rest of the fit to SCr was good. GFRs(t) varied more than what is desirable, however, the total variation was less then what is seen for other patients.

Figure 3.10: Patient 9 (fatality) data and fitted results. Time in hospital = 378 h.

Over the course of 394 hours, patient 10 (Figure3.11) received three total fluid types: IV, oral intake, and continual infusion. The rate of continual infusion was significant increased over the course of a day. This significant fluid increase corresponded to a significant, continual loss of interstitial fluid from t = 75 h to t = 180 h. The range of SCr values is from 9 mg L-1 to 93 mg L-1_{. The model did not properly fit UOT. From t = 100 h to t > 350 h, the urine} production rate was very minimal and the model was not capable of fitting this very small

increase. An underlying condition present in patient 10 not captured by the framework of the model is responsible for this missed ability to capture the results. Due to this, the model assumed severe dehydration and fit the SCr data with GFRs(t). At the same time point that SCr data began to increase, urine production significantly decreased (nearly nothing at all). This may infer the onset of anuria as the sharp increase in SCr would conclude highly minimal kidney function. A model capable of capturing the development of anuria would be capable of fitting this data set. GFRs(t) varied with SCr more than what would be desired. Overall, this fit indicates an opportunity to further develop this model.

Table 3.2: Patient data fit calculations SSE, MSE, and RMSE

Patient SSE SCr MSE SCr RMSE SCr SSE UOT MSE UOT RMSE UOT

# ( mg2 _L-2 _{) ( mg}2 _L-2 _{) ( mg L}-1 _{) ( L}2 _{) ( L}2 _{) ( L )} 1 22.75 1.75 1.32 38.05 0.67 0.81 2 0.18 0.01 0.11 3.61 0.36 0.60 3 84.27 3.83 1.96 6.82 0.11 0.33 4 516.11 19.11 4.37 1279.78 9.77 3.12 5 48.74 1.87 1.36 63.62 0.41 0.64 6 3.38 0.18 0.42 8.00 0.28 0.53 7 714.43 28.57 5.35 268.87 4.34 2.08 8 29.48 1.34 1.16 384.70 3.37 1.84 9 664.02 32.20 5.67 200.17 3.39 1.84 10 56.14 2.55 1.60 93.22 1.73 1.31

Table 3.3: Patient kc parameter fit results

Patient # kc,vol kc,min kc,max kc,slope 1 42.54 0.001 0.600 -1.5 2 55.20 0.015 0.600 -3.0 3 55.19 0.020 0.600 -3.0 4 25.82 0.02 1.200 -1.5 5 44.90 0.020 1.200 -1.5 6 50.86 0.012 0.898 -2.6 7 35.83 0.001 0.600 -3.0 8 55.19 0.001 0.600 -3.0 9 39.61 0.020 0.675 -1.5 10 55.16 0.014 0.652 -3.0