Page 183 - Functional impairment and cues for rehabilitation of head and neck cancer patients -

Decision curve analysis
Figure 4 represents the net benefit of applying the model for each risk threshold. The decision curve shows that the prediction model has net benefit over current practice which includes reactive feeding tube placement (treat none), for probability thresholds from 35–80%.
Figure 4 Decision curve analysis. Solid black line: Net benefit of risk prediction model relative to reactive feeding tube placement protocol in all patients with 95% confidence intervals. Dashed black line: Net benefit of providing all patients a prophylactic feeding tube or treat all with 95% confidence intervals (‘’All’’). Solid grey line: Reference, reactive feeding tube placement protocol in all patients or treat none (“None”). The decision curve shows that the prediction model has net benefit for risks from 35–80% over current practice which includes reactive feeding tube placement (treat none).
Influence of factors after (start of ) CRT
Timing of placement is associated with prolonged tube dependency. Patients who received a tube before or in the first week of CRT (n = 59) have a higher risk of prolonged tube dependency (OR 25.92; 95% CI 9.12–73.69, p < .001). Per day later after start of CRT, patients have a lower risk of prolonged tube dependency (OR 0.95; 95% CI 0.93–0.97, p < .001).
Patients who received a neck dissection within the first 90 days after CRT (n = 8) did not have a significantly higher risk of prolonged feeding tube dependency (OR 2.08; 95% CI 0.49–8.84, p = .322), however, group size was small.
7
Prediction model for tube feeding during CRT
181