RESUMO
BACKGROUND: Primary care chronic kidney disease (CKD) registers report widely varying prevalence within the UK. We examined the effects of laboratory ascertainment and adjusting for practice-level variables on the variation in CKD prevalence. We carried out an Ayrshire-wide laboratory database analysis of primary care practices (PCPs). METHODS: We analysed 54 PCPs with 313 639 registered patients aged ≥ 18. All patients with a low estimated glomerular filtration rate (<60 mL/min/1.73 m(2)) had their serum creatinine values extracted from 1st January 2009 to 31st March 2012. Individuals with CKD stage 3-5 were identified with an algorithm that confirmed chronicity. These data were linked to PCP attributes from Information Services Division, Scotland. Using laboratory-ascertained CKD prevalence, we examined whether adjusting for practice-level factors [socio-economic status (SES), rurality and patients to general practitioner ratio (PGR)] and patient-level factors (age, gender) explained some of the observed variation among PCPs. Individual and combined hierarchical multilinear regression models were used. RESULTS: Eighteen thousand two hundred and eighty-five (5.8%) had CKD stage 3-5 on 31 March 2011. SES, rurality and PGR predicted 39% (F(3,50) = 12.37, P < 0.001) of the variation in prevalence with SES exerting the most influence (25%). With the stepwise addition of explanatory variables, variation between practices fell from 3.9-fold using PCP register prevalence to laboratory ascertained (3.1-fold variation), with age and gender adjustment (further fall to 2.1-fold), and lastly to 1.8-fold variation with adjustment for SES. Funnel plots using these adjustments reduced the number of outliers outside of 3 SD from 15 to 7 to 6, and outliers between 2 and 3 SD by 16 to 13 to 5. CONCLUSIONS: Laboratory ascertainment is practicable, reduces variation and facilitates benchmarking. PCP attributes other than age and gender impact on prevalence. Over a third of variation in CKD prevalence among PCPs can be explained by rurality, PGR and especially SES even after age and gender stratification.