OUP user menu

Late referral for dialysis: improving the management of chronic renal disease

P. Roderick, C. Jones, C. Tomson, J. Mason
DOI: http://dx.doi.org/10.1093/qjmed/95.6.363 363-370 First published online: 1 June 2002


Background: Timely nephrological referral of patients with chronic renal failure (CRF) is important, but referral at a late stage of disease is common.

Aim: To investigate whether late referral of patients is avoidable, and where the missed opportunities lie.

Design: Prospective ascertainment of new cases and comprehensive review of pre‐end stage history.

Methods: Patients admitted to Bristol and Portsmouth renal units for chronic RRT between June 1997 and May 1998 were identified from computer databases. Data were collected from case notes and hospital records, and a self‐administered patient questionnaire. Late referral, defined as dialysis within 4 months of first referral to a dialysing nephrologist, was categorized by algorithm as unavoidable or avoidable.

Results: Of 250 patients, 96 (38%) were referred late. Forty‐three (45%) had definite avoidable reasons: 35 (37%) with raised serum creatinine for a median 3.7 years (IQR 1.5–8.2) before referral, and eight (8%) with risk factors for renal disease but scant assessment of renal function; 12/43 (31%) had a diagnosis of diabetic nephropathy. Late referred patients were less likely to receive standard renal therapies for chronic renal failure, were in a poorer clinical state at start of RRT, and more often required emergency dialysis, compared to patients referred early. Late referrals were as likely from a hospital as a primary care physician.

Discussion: A significant proportion of patients are avoidably referred to a dialysing renal unit at a very late stage. Guidelines on referral should be developed by nephrologists, primary and secondary care physicians, and patient groups, and further research is needed into the cost‐effectiveness of early referral strategies.


Renal replacement therapy (RRT) is a lifesaving treatment for end‐stage renal failure (ESRF). Provision has increased steadily in the UK, accounting for approximately 1.5% of NHS resources,1 and demand is predicted to double over the next 10–20 years.2 Prevention of ESRF is therefore critical. Evidence is accumulating that the progression of chronic renal disease to ESRF can, in some cases, be delayed, halted or even reversed by a range of interventions.3 Moreover, chronic renal failure (CRF) is associated with morbidity, such as renal bone disease and anaemia, which impairs patients' quality of life, but for which there are effective interventions.

Timely referral of patients with CRF to specialist nephrological care is needed to ensure the introduction of such measures early enough in the disease process to provide benefit. However, 30–40% of patients are referred to a nephrologist at a very late stage of renal disease,4–,7 and this is associated with a poorer clinical state at start of RRT,8 a worse prognosis9 and higher initial health care costs.10 Earlier referral is also important in preparing the patient adequately for dialysis, with education, counselling and the formation of dialysis access.

Few studies have attempted to determine the underlying reasons for late referral.4,9,,11 To find ways of improving practice, it is important to distinguish unavoidable late referral, (i.e. those who present for the first time with advanced renal failure, or those with rapid‐onset renal disease or an acute exacerbation of previously unknown chronic renal failure) from avoidable late referral (characterized by patients with clinical evidence of renal impairment that were not acted upon, or those at high risk of renal disease with missed opportunities for detection).

This study aimed to investigate to what extent late referral is avoidable, and to assess where the missed opportunities for earlier intervention lay.


All patients admitted to Bristol and Portsmouth renal units for chronic RRT between June 1997 and May 1998 were eligible. These are regional units with renal transplantation facilities, supplying a largely Caucasian, mixed urban population in the south of England. Potentially eligible patients were identified from unit computer databases that record details of all new acceptances for RRT. Patients with recovered acute renal failure, those transferred in from other units already on dialysis, and returns to dialysis from failed transplants were excluded.

Data were collected from renal unit and primary care case notes, renal unit computer records, referring hospital records and by self‐administered questionnaires to the patients. Relevant data items were agreed by an expert panel, including a geriatrician, a primary care physician, a diabetologist, two nephrologists involved in dialysis, and one general physician with an interest in nephrology.

Dialysis within 4 months of first referral to a dialysing nephrologist was used as the definition of late referral, as this was considered the absolute minimum time needed for dialysis access formation and preparation of the patient. History of chronic renal failure was defined as persistently raised serum creatinine >150 μmol/l and/or a diagnosis of chronic renal failure recorded in the medical notes.

Patients’ co‐morbidity was classified using the risk index of Khan et al.12

A ‘patient timeline’ (Figure 1), was constructed for each late referred patient. Serum creatinine and blood pressure results prior to referral were plotted against clinical history such as urinalysis, prescribed anti‐hypertensive medication, hospital referrals and admissions, renal investigations and diagnoses. Patients were then categorized independently by each of the authors, according to an algorithm (Figure 2), with borderline cases decided by consensus. The timelines were also used to elicit missed opportunities for early referral and intervention. In cases where the serum creatinine rose over time, reciprocal creatinine plots were generated to assess whether this was linear. The categories used were:

Unavoidable late referral

(i) Acute irreversible renal failure

A sudden rise in serum creatinine with no previous evidence of renal failure or missed opportunities for detection, and a diagnosis not consistent with chronic renal failure.

Figure 1. 

Example of patient timeline used for classification and investigation of missed opportunities. Male patient, diabetic nephropathy. Raised serum creatinine not acted upon. Scr, serum creatinine; Sbp, systolic blood pressure; Dbp, diastolic blood pressure.

Figure 2. 

Characterization of patients.

(ii) Acute on chronic renal failure

A sudden rise in serum creatinine suggesting an additional renal insult (e.g. nephrotoxic agent), either on pre‐existing stable CRF, or probable CRF deduced post referral (e.g. small kidneys on ultrasound).

(iii) Late presenters

Patient asymptomatic until final presentation in ESRF with no previous evidence of renal failure or of missed opportunities for detection.

(iv) Non‐attenders

Frequent hiatus in follow‐up with evidence of multiple missed follow‐ups.

Avoidable late referral

(i) Evidence of renal disease

Haematuria, proteinuria, or raised serum creatinine, with no referral.

(ii) Missed opportunities for detection of renal damage

Urinalysis or serum creatinine not tested or very infrequently tested in those at high risk of renal disease, (i.e. patients with diabetes, hypertension, a family history of renal impairment, or previous urinary outflow obstruction).

Statistical analysis

Statistical analysis included the comparison of patient group characteristics using tests appropriate to their data and distribution (χ2, t‐test, Mann‐Whitney). Logistic regression analysis was used to identify independent predictors of death. p values <0.05 were accepted as significant. All calculations were carried out using SPSS (version 9.0). Ethical approval was granted by Portsmouth and South East Hampshire Health Commission Research Ethics Committee, and United Bristol Healthcare NHS Trust Research Ethics Committee.


We identified 253 eligible patients in the study period; two died before giving consent, and one had no medical records available, thus 250 patients entered the study. Renal unit information was available for all patients. Overall, 86% (216) of primary care proformas were completed, with unavailable data due to no patient consent given (14), records being unavailable (6), or general practice inaccessible for data collection (14) (i.e. Channel Islands). Sixty‐two patients were referred direct from the primary care physician, and for 168 (84%), referring hospital proformas were completed. Overall, 90% (225) of patient questionnaires were completed; missing data were due to patients being unwilling or unable to complete the questionnaire.

Table 1 shows the distribution of patients categorised as early referrals, avoidable or unavoidable late referrals. Of the 96 (38%) patients referred late, 49 (51%) were judged to be unavoidable and 43 (49%) avoidable; 4 (4%) were unclassifiable due to lack of data. 35 (81%) of the avoidably late referred had known progressive renal damage (i.e. serum creatinine persistently >150 μmol/l and rising), and 8 (19%) had missed opportunities for earlier detection of renal impairment. These included no measurements of renal function in one patient with prior urinary outflow obstruction, and seven patients with sustained hypertension (>140/80 mmHg), three of whom also had persistent proteinuria (>2+ on dipstick).

Compared to the early referred group the avoidable late referrals were as likely to be referred by a hospital as a primary care physician (Table 2). There was no significant difference in their age or co‐morbidity. The avoidable group had significantly adverse markers of CRF associated co‐morbidity at start of RRT, including lower albumin (35 vs. 38 g/l, p<0.05), lower calcium (2.1 vs. 2.6 mmol/l, p<0.001) and lower haemoglobin levels (9.2 vs. 11.2 g/dl, p<0.001).

The avoidably late referred were significantly less likely than the early referred to have received standard renal therapies (Table 3), including erythropoietin (p<0.01) and vitamin D supplements (p<0.01), and were less likely to be taking aspirin (p<0.05), a recognized secondary prevention measure for patients with CRF and cardiovascular disease. They had less permanent dialysis access in place at start of RRT (p<0.01) and had a significantly longer initial hospital stay (p<0.01). There was little difference between the early and avoidably late referred in the number of patients dying within the first 90 days [11 (9%) vs. 2 (5%) p=NS], and 6 months [19 (16%) vs. 11 (28%) p=0.05]. Logistic regression adjusting for age and co‐morbidity showed that being avoidably late referred was not significantly associated with mortality at 6 months (OR 2.15, 95%CI 0.91–5.08, p=0.08). Mean blood pressures 1, 3 and 5 years prior to referral did not differ significantly for all groups, but none met the British Hypertension Society recommendations13 for blood pressure control current at that time (<140/80 mmHg).

The main category in the avoidable group was that of patients (n=35) with rising serum creatinine that did not prompt referral. Their median duration of known chronic renal failure was 3.7 years (IQR 1.5–8.2), yet they were referred a median 25 days (IQR 3–65 days) prior to requiring RRT. The median (IQR) number of serum creatinine tests prior to referral was 18 (11–31). The commonest cause of renal disease was diabetic nephropathy (n=11, 31%). Nine (26%) of the 35 patients were under the care of a primary‐care physician with no record of hospital contact, eight (23%) had primarily primary care with occasional secondary input and 18 (51%) were predominantly seen by a hospital physician. They were under the care of these physicians for a median period of 6 years (IQR 3–9). Documented dipstick urinary abnormalities were present at first diagnosis of CRF in 68% of the 35 patients (haematuria, proteinuria or both). Renal ultrasound, a standard first line investigation, was only used in the diagnosis in 12 (34%) patients prior to referral.

Table 4 shows the characteristics of the patients in whom there were other missed opportunities for earlier detection. Those with hypertension were mostly under the care of a primary care physician, for approximately 3 years (median, IQR 1–6). They had known hypertension for 6 years (median, IQR 5–21) and were referred to a nephrologist only 10 days prior to starting RRT.

View this table:
Table 1 

Avoidability of late presentation

Early referralsEarly referred154 (62%)
Avoidable late referralsRenal damage ignored 35 (14%)
Missed opportunities for detection: proteinuria   3 (1%)
Missed opportunities for detection: hypertension   4 (2%)
Missed opportunities for detection: urinary obstruction   1 (1%)
Total 43 (17%)
Unavoidable late referralsLate presenters in ESRF 22 (9%)
Multiple DNA   4 (2%)
Acute renal failure 11 (4%)
Acute on chronic renal failure 12 (5%)
Total 49 (20%)
UnknownData unavailable to categorize   4 (2%)
Total250 (100%)
View this table:
Table 2 

Baseline characteristics of early, avoidable and unavoidable late referral groups

CharacteristicEarly referral (n=154)Avoidable late referral (n=43)Unavoidable late referral (n=49)Total (n=250)
Mean (SE) age (years)55 (1)60 (2)62 (2) 57 (1)
Males (%)85 (55)28 (65)31 (63)146 (58)
Khan Index at RRT n (%)
High risk31 (20) 8 (19)15 (31) 54 (22)
Medium risk65 (42)24 (56)21 (43)110 (44)
Low risk58 (38)11 (26)13 (27) 82 (33)
Referral source n (%)
General physician*54 (35)29 (67)21 (43)104 (42)
General practitioner*56 (36) 4 (9) 7 (14) 68 (27)
Urologist 9 (6) 2 (5) 3 (6) 14 (6)
Diabetologist17 (11) 2 (5) 1 (2) 20 (8)
Other18 (9) 6 (14)17 (35) 41 (18)
Diagnosis group n (%)
CRF, unknown cause35 (23)11 (26)14 (29) 60 (25)
Glomerulonephritis31 (20) 6 (14) 5 (10) 42 (17)
Pyelonephritis13 (9) 3 (7) 3 (6) 19 (8)
Polycystic kidney disease15 (9) 3 (7) 2 (4) 20 (8)
Renovascular disease 7 (5) 3 (7) 6 (13) 16 (7)
Diabetic nephropathy35 (24)12 (28) 1 (2) 48 (20)
Other cause18 (10) 5 (10)18 (37) 41 (13)
  • *p<0.001, comparing early referrals with avoidable late referrals.

View this table:
Table 3 

Prior treatment, blood pressure control and outcomes of early, avoidable and unavoidable late referral groups

Early referral (n=154)Avoidable late referral (n=43)Unavoidable late referral (n=49)Total (n=250)
Prior treatment at RRT n (%)
Aspirin‡ 37 (25)   5 (12)   7 (15) 49 (21)
Vitamin D† 58 (39)   8 (19)   4 (8) 70 (29)
Phosphate binders 79 (52) 17 (40) 10 (20)106 (43)
Sodium bicarbonate 23 (15)   2 (5)   5 (10) 30 (12)
Lipid‐lowering medication 30 (20)   6 (15)   4 (8) 40 (17)
Erythropoeitin† 67 (44)   9 (23)   4 (8) 80 (34)
Mean (SE) blood pressure pre‐referral (mmHg)
1 year pre‐referral159/88152/85148/82156/86
3 years pre‐referral156/88158/88149/87155/88
5 years pre‐referral153/86151/86150/87152/87
Outcomes n (%)
Haemodialysis as first RRT 90 (60) 29 (67) 44 (92)163 (67)
Permanent access at RRT† 99 (69) 20 (47)   8 (17)127 (55)
Mean (SE) initial hospital stay (days)† 18 (2) 36 (11) 39 (11) 27 (6)
Dead at 90 days 11 (9)   2 (5) 10 (21) 23 (11)
Dead at 6 months‡ 19 (16) 11 (28) 17 (35) 47 (23)
  • p<0.01; ‡p<0.05, early vs. avoidable late referrals.

View this table:
Table 4 

Characteristics of patients with missed opportunities for detection

Missed opportunityAge groupSexHypertension* (days)Proteinuria* (days)Duration of CRF (days)PhysicianDiagnosis
Proteinuria65+F16851530 904DiabetologistDiabetic nephropathy
65+F 928 457 679GPRenovascular
45–64M 257 334   71UrologistUnknown
Hypertension45–64M2121N/A 196UrologistUnknown
65+F7560N/A   27GPUnknown
45–64M1747N/A    0GPUnknown
65+F1774N/A   20GPUnknown
Urinary obstruction16–44MN/AN/A1668OtherCongenital
  • *Time from diagnosis to first raised serum creatinine.


In this study, while the late referral of patients requiring dialysis was common, almost 50% of late referrals were unavoidable due to the sudden or insidious nature of their renal failure. In these circumstances, there was little that the referring physician could have done to initiate earlier treatment. Nevertheless, one in six of all those starting RRT were avoidably late‐referred, and might have benefited from earlier intervention. These patients started RRT in a poorer clinical state, were less likely to have received interventions aimed at slowing progression or treating CRF‐associated morbidity, were less well prepared for RRT, and had a poorer outcome on RRT. We found no evidence that age9 or co‐morbidity14 were significantly associated with avoidable late referral, as has been found in some5,12,,15 but not all9 previous studies, although none analysed an avoidable category separately.

We devised our classification system to identify avoidable late referral. Other late referral studies have acknowledged the presence of ‘unavoidable’ late referrals,9,4,,11 but due to lack of pre‐referral data and small sample size, they were unable to categorize accurately the path of the patient from first renal impairment to referral. Two more recent studies have categorized the flow of renal patients into dialysis programmes15 and investigated the factors associated with early death on dialysis.16 The first did not divide the sample by late presentation/non‐attendance; the second classified patients as planned or unplanned presentation according to the presence or absence of permanent dialysis access at start of RRT. This definition differed from ours, making comparison difficult, but they did find a much higher proportion of both acute on chronic renal failure (12% vs. 5%) and acute renal failure not recovered (11% vs. 4%). This may be due to the definitions used, as we classed a long period of progressive chronic renal failure not under a nephrologist as a missed opportunity for earlier referral, whether or not the final presentation was due to a more rapid decline in renal function.

The key group who could have benefited from earlier intervention were the 35 (14%) with a progressive rise in serum creatinine that was not acted upon. This is a disappointing finding in light of the UK Renal Association17 recommendation that all patients with a serum creatinine >150 μmol/l be referred for nephrology assessment. However these guidelines were written primarily for renal physicians, and have not been widely disseminated to physicians of other specialties or primary care. Some 31% had ESRF due to diabetes and 6% due to hypertension, both chronic conditions requiring proactive care.18 This highlights the importance of close co‐operation with diabetologists, and guidelines for referral of patients with type 2 diabetes who are often managed in primary care.

The other (smaller) avoidable group were those in whom renal impairment should have been considered. Some patients had hypertension19 with or without abnormal urinalysis, or previous urinary obstruction,20 both risk factors for progressive nephropathy, and yet they had either no or very infrequent renal investigations. This accords with other evidence of underassessment of renal function in hypertensive patients; Kissmeyer21 found that only 53% of the primary care records of 2561 hypertensive patients had evidence of an annual creatinine test.

In contrast to Sacks20 et al., we found only one of 96 late referrals (<1%) at high risk of renal disease due to previous urinary obstruction that was not acted upon compared to 4/19 (21%), and none of our patients had previous prostatectomy.

Amongst the unavoidably late‐referred, there were very few (4, 2%) not attending for nephrological follow‐up. This suggests that the current services are both accessible and acceptable to patients.

Why are patients not referred earlier? A study of referring physicians in the USA found that there was no difference in demographic or professional characteristics in those whom referred patients late or early, but a perceived lack of training and guidelines in the referral of patients with renal disease, and poor communication with nephrologists were factors.22 Our late referrals came from a broad spectrum of hospital and primary care physicians. This might reflect a failure to recognize the importance of a rising serum creatinine and the potential benefits of nephrological assessment, or ambivalence about the benefits of such assesment and the patient's suitability for RRT.

The strengths of the study were its high completeness of data on prior history, and its inclusion of more than one renal unit. It was limited by the potential under‐ascertainment of clinical variables, especially tests with normal results. Also the findings should not be generalized to areas with large ethnic minority populations at higher risk of ESRF. While we have arbitrarily taken late referral to be 4 months, there are arguments a longer period. The Canadian Society of Nephrology recommends23 12 months to adequately prepare a patient for dialysis, while Jungers24 suggests that 3 years is needed to influence the increased cardiovascular risk associated with renal disease.

What are the implications? Late referral should be monitored and where possible distinguished as unavoidable or avoidable by national Renal Registries. Whilst the Renal Association17 recommends the referral of all patients with serum creatinine >150 μmol/l, the majority of these patients will not progress to ESRF, and mild chronic renal failure may be more important as a risk marker for cardiovascular disease.25 Moreover, mild chronic renal failure is common26 and it is doubtful whether current UK nephrology services could manage comprehensive early referral. There is a need for guidelines on referral, developed by nephrologists, primary and secondary care physicians and patient groups. Better evidence is required of the effectiveness and costs of earlier referral strategies and methods of identifying those most likely to progress to end‐stage renal failure.


We are very grateful to our study nurses Linda Bourton and Sue Garland for their hard work and dedication in collecting the data. We would like to thank the renal units, general practices and referring hospitals staff for facilitating the data collection and the patients themselves for taking part. The study was funded by a Research Grant from the South East NHS Executive Research & Development Department, UK. CJ was part funded by Baxter Healthcare Ltd. No conflict of interest was evident. PR conceived and led the project and oversaw the writing of the paper. He is guarantor. CJ undertook the analysis and wrote the first draft of the paper. JM and CT provided clinical advice throughout, oversaw data collection at each unit, and contributed to the paper.


  • Address correspondence to Dr P. Roderick, Health Care Research Unit, Level B, South Academic Block, Southampton General Hospital, Southampton SO16 6YD. e‐mail: pjrsoton.ac.uk


View Abstract