The Peritoneal Dialysis Surprise Question and Technique Survival: Are you surprised?

Patients with advanced chronic kidney disease often find themselves in a situation where a pre-emptive kidney transplant is not possible and so they need to be initiated on dialysis. There has been a global push towards increasing the utilisation and prevalence of home dialysis (peritoneal dialysis (PD) and home haemodialysis (HHD)) through increasing patient and provider education, expanding home dialysis manufacturing capabilities and awareness, as well as the implementation of PD-first policies in some countries. However, the success of the dialysis modality selection process should not be simply measured by an increase in the proportion of incident dialysis patients on PD and HHD. Technique persistence is the key here, which is demonstrated through the prevalent number of patients on PD and HHD.

Dialysis initiation is associated with significant morbidity and mortality. Approximately 25% of maintenance dialysis patients have unipolar major depression or other depressive syndromes, ¹ resulting in increased hospitalisation rates, ² withdrawal from dialysis ³ and all-cause mortality. ⁴ In fact, the US Renal Data System (USRDS) reports a 1-year all-cause mortality of 15–20% in incident dialysis patients, ⁵ with some reporting rates as high as 34%. ⁶ Moreover, it has been demonstrated that the first 2 weeks of chronic dialysis are associated with 2.72-fold risk of death and 1.95-fold risk of hospitalisation compared to patients who survived their first year of chronic dialysis. ⁷

While many of these outcomes are multifactorial in nature, owing to patient comorbidities, age, prescription patterns as well as dietary and treatment non-compliance, one thing that is tangible is modality selection. Approximately 45–57% of patients chose a home dialysis modality after attending a pre-dialysis education programme. ^8,9 The modality selection process is one in which patient preferences take centre stage and it is the provider’s role to guide and support patient informed decision-making. Assessing the appropriateness of dialysis modality selection should be based on three criteria: eligibility for modality of choice, degree of information received about modalities and patient perception of the shared decision-making process. ¹⁰ In a cross-sectional study of incident dialysis patients, 22% of the 141 study participants were deemed to have a potentially inappropriate dialysis modality selection process. ¹⁰ It is important to note that only 1 of those 31 patients was started on PD, indicating a potentially more diligent modality selection process for patients on home modalities.

Absolute contraindications to PD initiation include an insufficiently clean environment to perform exchanges, inadequate cognitive/physical ability by the patient and caregiver to perform PD or lack of suitable peritoneal cavity due to extensive scarring or adhesion – something which can only be definitively determined laparoscopically. ¹¹ Relative contraindications include morbid obesity, polycystic kidney disease, presence of an ostomy, cirrhosis and severe patient cognitive or physical impairment – all of which have potential solutions. ¹¹

Modality selection is an integral and delicate process, the success of which is measured by technique failure rates. Some studies report kidney transplantation as a cause of ‘technique failure’; however, that is a positive outcome and should not be interpreted as ‘failure’. Multiple attempts at evaluating early PD technique failure rates adopted varying definitions based on timeline (6 months vs. 12 months), definition of technique failure (transfer to HD vs. cessation of PD due to PD-related issues) and whether or not to include death as a cause of technique failure. ¹² Reported technique failure rates range from 12.7% to 24.8% within 12 months of initiation. ^{13 –15} Technique failure and transitioning to in-centre haemodialysis (HD) is associated with a considerable emotional, physical and economic toll. The majority of patients who discontinue PD are converted to in-centre HD. ¹⁶ This is associated with high rates of acute care encounters (emergency department encounters, observation stays and hospital admissions) and healthcare expenditures. ¹⁶

To date, there have been several prediction models that help identify PD patients at risk of transfer to HD ^{17 –19} with varying degrees of success. Furthermore, they have differing definitions of transfer to HD and use different variables to predict transfer. Conceptually, this is an important tool that can help providers guide patients in the dialysis modality selection process. In this issue of Peritoneal Dialysis International, Hussein et al. ²⁰ set out to utilise a modified version of the point of care Surprise Question used by palliative care providers as a screening tool to identify patients who providers believe are at risk of transferring to HD within 6 months. In this prospective observational study, nephrologists and registered nurses in a non-profit dialysis organisation in the United States were asked the PD Surprise Question (PDSQ) ‘Would you be surprised if this patient transferred to HD in the next 6 months?’ All prevalent PD patients in December 2020 were included. The primary outcome was transfer to HD, with death, hospitalisation and peritonitis as secondary outcomes.

PDSQ responses were obtained for a total of 1275 patients. There were 1123 patients in the registered nurse (RN) cohort, 692 patients in the medical doctor (MD) cohort and a total of 616 patients had responses from both nurses and nephrologists (RN-MD cohort). During the study follow-up period, 4.3% of the total patient cohort transferred to HD. Using concordance statistics, the RN cohort had C-statistic values of 0.77 (60 days), 0.64 (90 days) and 0.63 (170 days) for transfer to HD. The MD cohort had C-statistic values of 0.82 (60 days), 0.65 (90 days) and 0.68 (170 days). The RN-MD cohort had C-statistic values of 0.85 (60 days), 0.61 (90 days) and 0.68 (170 days). All three cohorts achieved high specificity values (≥83.4%), whereas the greater sensitivity values were seen at the 60-day mark; 68.5% (RN cohort), 80.4% (MD cohort) and 79.1% (RN-MD cohort). There was a sharp decline in sensitivity values in subsequent follow-up days, falling to the range of 30.8–51.1% across all three cohorts. Patients in the high-risk group were more likely to experience death and hospitalisation during the study follow-up period than low-risk patients. Interestingly, peritonitis rates were similar between both groups.

In general, potential causes for PD technique failure can be divided into three main categories: modality-related, patient-related and system-related. ²¹ Examples of modality-related issues include peritonitis, tunnel and exit site infections, ultrafiltration failure and catheter problems. Patient-related issues include burnout, malnutrition, abdominal surgeries, stroke and severe illness. System-related issues include provider expertise, lack of infrastructure, lack of modality education, physician reimbursement and dialysis facility ownership. ²¹ This is why predicting PD technique failure is a difficult task. Answering one question, such as the PDSQ, relies on the providers’ ability to think through all these factors and determine what they believe would be a patient’s ultimate modality outcome.

The use of the Surprise Question in palliative care, ‘Would you be surprised if this patient died within the next x months?’ has yielded variable results. In a meta-analysis of 26 papers, ²² the pooled accuracy level of the surprise question was 74.8% (95% CI 68.6–80.5%). Overall, there was a wide degree of accuracy, ranging from poor to reasonable. This helps set the stage for interpreting the PDSQ study outcomes and set thresholds for expectations.

In Hussein et al.’s study, there is a drop-off in the C-statistic and sensitivity values from the 60-day mark to the subsequent days of analysis. Given its dependence on sample size and the low event rate, the readers should be very cautious when interpreting the results of the positive and negative predictive values. C-statistic values >0.7 indicate a good model, whereas values >0.8 indicate a strong model. Values in this range are only seen across all three cohorts at the 60-day mark. Similarly, sensitivity values are >50% across all three cohorts only at the 60-day mark. In this case, the sensitivity of the PDSQ is its ability to correctly classify an individual as high risk for PD technique failure. This leads me to conclude that given the results of this study, the PDSQ is a potentially useful tool for identifying high-risk patients for PD technique failure within 60 days.

While peritonitis is one of the leading causes of PD technique failure, the similar rates seen in the high-risk and low-risk groups in Hussein et al.’s study are most likely due to a combination of two things: (1) the primary outcome of the study was transfer to HD, which is only one component of PD technique failure, and (2) this further demonstrates the multitude of factors that play a role in modality survival as mentioned earlier.

The drop-off in the predictability of the PDSQ after day 60 demonstrates just how dynamic the dialysis process is. The interplay of patient comorbidities, personal circumstances, dialysis-related issues and system-related issues all play a role in the unpredictability of patient outcomes beyond that point. In a time-dependent analysis of PD technique failure and mortality, Kolesnyk et al. ²³ demonstrated this interplay clearly. Outcomes in this study included transfer to HD, death, transplant, infections, catheter complications, abdominal complications and psychosocial/unknown. As discussed earlier, safer access planning and transitions to HD are needed for patients on PD. The PDSQ adds to the arsenal of tools at clinicians’ disposal to help make this determination. Although it does not allow enough time for arteriovenous access creation and maturation, at least it would give patients and providers an opportunity to avoid patients ‘crashing’ into HD in an unplanned, urgent manner.