Novel use of a combined primary and secondary care data resource: investigation of inequalities in chronic kidney disease identification in the UK Hampshire Health Record

BACKGROUND: Chronic kidney disease (CKD) is classified into five stages from 1 (mild) to 5 (severe), and is identified from estimated glomerular filtration rate (eGFR). The 2010 Health Survey for England estimated that 6% of men and 7% of women have stage 3–5 CKD, increasing with age to 29% of men and 35% of women older than 75 years. CKD is an example of a chronic condition managed in primary care under the UK Quality and Outcomes Framework (QOF). People with CKD have higher morbidity and mortality than do the general population through greater risk of cardiovascular disease, progression to end-stage renal failure, and other complications. There is evidence that the true prevalence could be greater than that in the QOF-registered population, suggesting under-recognition in primary care. People on QOF disease registers are treated with standardised disease management protocols, the absence of which might result in poor outcomes. We aimed to identify inequalities in disease identification in people with CKD 3–5. METHODS: We constructed a retrospective cohort of people with incident disease stage 3–5 between 2008 and 2011 from the UK Hampshire Health Record (HHR), a database combining anonymised routine data from both primary and secondary care that receives data from 133 general practices (total registered population about 1·1 million). The HHR also contains laboratory data, allowing biochemical classification of CKD with agreed definitions, rather than relying on QOF coding. New CKD was defined as having two values of eGFR lower than 60 mL/min per 1·73 m² at least 3 months apart (and previous eGFR >60 mL/min per 1·73 m² or no previous eGFR value). QOF-registered CKD was identified by relevant Read codes. Socioeconomic status was defined by index of multiple deprivation (IMD), categorised into tertiles. The main outcome of our study was appropriate identification of CKD in primary care as recorded for the QOF register. We compared QOF identification of CKD with study-defined disease by sociodemographic and comorbidity variables using univariate and multivariable logistic regression. FINDINGS: 11 057 people with incident CKD 3–5 were identified from the database. Of this population, 4769 (43%) were male, 6288 (57%) female, 9909 (90%) were older than 60 years, 1766 (16%) were in the most deprived tertile of deprivation, 6306 (57%) had a diagnosis of hypertension, and 1787 (16%) had diabetes. 3646 (33%) were on QOF CKD registers and 7411 (67%) were not. On univariate analysis, there was no difference in QOF registration status by sex. Younger people (odds ratio 2·03 [95% CI 1·74–2·36] for age <60 years compared with >80 years), and people in lower socioeconomic groups (1·14 [1·02–1·28] for most deprived IMD tertile), had lower odds of being QOF CKD-registered. People who also had diabetes or hypertension had greater odds of being QOF CKD-registered (1·38 [1·24–1·53], and 2·39 [2·19–2·60], respectively). In multivariate analysis, adjusted for age, sex, IMD, diabetes, and hypertension, younger people remained less likely (1·62 [1·39–1·90] for age <60 years compared with >80 years) and people with comorbidities remained more likely to be on a QOF CKD register (for diabetes, 1·23 [1·10–1·37]; for hypertension, 2·25 [2·06–2·45]). INTERPRETATION: This study has shown important discrepancies between QOF-registered and non-registered populations with CKD. It has also shown the potential for combined health records such as HHR to provide valuable information for population health analyses such as variation by sociodemographic variables and comorbidity. Independent identification of chronic disease might contribute to improved CKD management in primary care and reduction of health inequalities. A limitation of the study was the use of routine data. For example, reliance on registration by general practitioners of diabetes and hypertension could have resulted in missing exposure data. New cases in this dataset might have had previous CKD if the patient moved from a non-HHR area during the study period. FUNDING: None.