Association between non-registration of chronic kidney disease and mortality and cardiovascular outcome: a time-to-event analysis of retrospective primary care data

Principal findings

This study showed that patients with a properly registered diagnosis die less quickly than non-registered ones. However, according to our results, these patients did not appear to have a lower risk of developing CVD. Besides, patients in stages 3B and 4 with a registered diagnosis had much better mortality survival rates compared with the non-registered ones. The association of non-registration and CVD was less clear in the different CKD stages.

Patients with CKD and hypercholesterolaemia were shown to be less associated with CVD and mortality, although the result was not statistically significant. In contrast, hypertensive CKD patients appeared to have a higher risk of CVD, but a lower risk of mortality. Patients with a history of CVD seemed to have a lower risk of new events, but a higher risk of dying.

Context of the results

Non-registration appears to be associated with all-cause mortality. However, the CaReMe CKD study recently showed that the rates of cardiovascular and all-cause death were 31%–49% higher in registered CKD patients than in measured CKD patients, which could not be confirmed in our study.25 Few researches have been conducted in this regard, making it difficult to compare. We must note that non-registration may be a risk factor for mortality comparable to diabetes, but outweighed by age and stage of CKD by far. An association was found, but causality was not investigated. It is unclear whether better registration will lead to a better outcome, so this should be a topic for further research.

The key research question of our results is what caused patients to die. Previous research showed that a reduced kidney function predicts both cardiovascular and non-cardiovascular mortality due to pulmonary disease, infection, cancer and other causes.26–28 The association between a reduced eGFR and the increased risk of cardiovascular events and hospitalisation was also found.9 12 29 Surprisingly, the non-registered group in our study did not have a higher risk of CVD than the registered. It is unclear why no association was found. Possibly, this group died more frequently as a result of non-CVD. On the other hand, non-registration probably extends beyond renal insufficiency and also occurs with other pathologies. Mata-Cases et al reported non-registration of diabetes mellitus in Spanish primary healthcare.30 Cardiovascular diagnosis may also be non-registered, so that no association with non-registration could be found.31

A second important question remains why the difference in mortality outcome was found between registered and non-registered patients. Is the root of the problem with the GP or the patient? Our previous research showed that there were small differences between registered and non-registered patients at baseline.15 Hypertension was more frequently present in the registered (64.4% of the registered population) compared with the non-registered (51.7% of the non-registered population). Similar results were found for type 2 diabetes (33.1%) of the registered compared with 28.2% of the non-registered). Small differences were also noticeable in the use of ACE-inhibitor (ACE-I) or angiotensin receptor blocker (ARB) (52.6% among the registered compared with 46.3% of the non-registered).15 However, these small differences do not seem to provide an adequate explanation for the difference in mortality, partly in view of the result that the non-registered group had no higher risk of CVD.

Subsequently, the follow-up of these patients should be assessed. A possible explanation for the difference in mortality between registered and non-registered groups could be that less attention was paid while prescribing and dispensing nephrotoxic (over-the-counter) medication by the GP and pharmacist, resulting in further deterioration of kidney function. There may have been less attention to the CVR factors associated with impaired renal function. In that case, we also would have expected an increase in CVD, unless, as previously described, it concerns a problem of global non-registration. On the other hand, the responsibility of the patient in the follow-up of the disease must be brought to attention. Possibly, the non-registered group contained a large proportion of patients who were not adherent to follow-up and therapy, as a result of which some did not or belatedly encountered problems. So, it is becoming increasingly important to examine these hypotheses and to involve the patient in his care and to find out what view he has in this regard.32 33 Moreover, it seems useful to investigate why the diagnosis was not registered in the EHR. Based on these results, the problem of non-registration could be addressed.

According to our research results, hypertension in CKD patients would be a risk factor in the development of CVD, although a protective factor in the development of all-cause mortality. Though, we know from previous research that hypertension is a risk factor for the development of CVD and premature death.34–36 The reasons for this difference are unclear. We need to consider the effect of antihypertensive medication on this outcome, since 48% of the patients took an ACE-I or an ARB.15 The beneficial effect of these drugs on cardiovascular events and all-cause mortality has been confirmed in the past.37 Ettehad et al described that in patients with CKD, smaller risk reductions in cardiovascular events were seen as a result of antihypertensive medication than in patients without CKD.38 However, we should also keep in mind that there may be non-registration of hypertension and SBP.

Additionally, it is surprising that hypercholesterolaemia and total cholesterol do not show a higher risk on CVD and mortality, since this is a proven risk factor for CVD.22 39 De Nicola et al showed that the CVR increases linearly with higher LDL in non-dialysis CKD patients.39 However, this result was not significant and may be explained by the use of lipid lowering medication, as 45% of patients were on this medication at baseline.15 After all, Fabbian et al determined that statins are an effective treatment in CKD patients, especially in the early stages of the disease.40 A history of CVD appears to be a protective factor in the development of new CVD, of which a properly adjusted therapy can be the reason (secondary prevention).40 41 In addition, we know that CKD is associated with adverse outcomes in those with existing CVD, which includes increased mortality after an acute coronary syndrome.42–44

In our previous work, we found that the majority of patients with renal insufficiency were in stage 3, with a higher proportion registered in stage 5 (75.7% registered) compared with stage 3A (22.9% registered).15 However, this study showed major differences in survival rates between registered and non-registered patients in both the earlier (3B) and further stages (4 and 5) of renal failure. The importance of early detection has been described many times in the past.17 45 This research must therefore be a plea for early detection of CKD and registration of the diagnostic code in the EHR. Good mutual communication between GP and nephrologist through referral letters and clear consultation reports can contribute to this. A solution to detect non-registered patients can be found in an Audit-& Feedback system, since this has proven to be effective and to have added value in primary care.46 47


There were some limitations to note. First, we did not take the presence of proteinuria into account in our CKD population. Mainly due to the lack of data on proteinuria, which brings us straight to the problem of non-detection of proteinuria in the Flemish general practice.

Subsequently, the study used healthcare data which may under-represent the healthy and asymptomatic that do not seek healthcare. The data of care refusers were included in the research results.

Although the patient population is representative for the Flemish population, registering GPs are not representative for the GP population. It is a selected group of high-quality registering practitioners that use a specific EHR. This selection bias of GPs could eventually have an influence on some process parameters in the follow-up of patients.16 In addition, data collected in a real-world setting may lack information on specific covariates and laboratory investigations. Lab results from the hospital and specialists are automatically entered into the EHR, but their diagnoses are not. The large proportion of missingness is a limitation as well. We used multiple imputations to fill in this missingness (see the Method section).

This post was originally published on