Healthcare cost analysis was performed on 864 patients, i.e. 884 patients recruited, net of 16 patients from one centre that had not completed follow-up and 4 patients whose costs were not filled in on the e-CRF. Overall, 586 patients completed the follow-up, whereas 278 patients interrupted the study before the last visit (32.2%; ranging from 51.4% in Veneto and 18% in Apulia and Basilicata) (Table 1). Among the dropouts, 36.0% of patients began dialysis treatment, 34.5% were lost to follow-up and 25.2% died. The study population mainly consisted of males (59.7%), and patients over 65 years of age accounted for 63.2% of the study population (18.6% were over 80 years of age); mean age was 66.3 years (± 14.6 years). At the time of enrolment, mean time from diagnosis was 7.0 years (± 7.9 years) and only 0.7% of patients were at their first nephrologist visit (data not reported in Table 1). Patients with a starting GFR level in stages 1, 2, 3, 4 and 5 totalled 68, 156, 355 (167 in stage 3a and 188 in stage 3b), 206, and 79, respectively. Hypertension, hypercholesterolemia and diabetes were the most frequent comorbidities, with 89.1% of patients experiencing at least one of these comorbidities and 107 patients (12.4%) being affected by all three.
Mean annual total (healthcare) cost per patient equalled €2723 (Fig. 1). All differences between values (status of patients, age, disease progression, comorbidities, starting GFR level, proteinuria at recruitment, and region) were found to be statistically significant according to the Kruskal–Wallis test, apart from the difference in unit costs between the male and female populations.
The unit costs of patients who dropped out were almost double those of patients who completed the follow-up. Dropped-out patients starting dialysis and dead patients during the follow-up period showed the highest mean annual cost, confirming that proximity to dialysis and death raises costs.
Disease progression produces an increase in costs. The higher the starting CKD stage, the higher the healthcare cost per patient. The largest difference between unit costs was found when comparing stage 3b patients with stage 3a patients (+ 56%) and stage 4 patients with stage 3b patients (+ 62%) (Table 2). Stage 3 accounted for more than 40% of patients.
Proteinuria also seems an important explanatory variable of annual cost per patient. Costs for patients without a proteinuria at the recruiting date were 25% less than healthcare costs for patients with proteinuria (Table 2).
Patients with CKD often have other diseases, with diabetes, dyslipidaemia and hypertension being the most important and frequent. Hypertension seems to have a larger impact on unit costs; patients in whom renal failure is associated with hypertension alone and hypertension with diabetes show higher costs. However, not all comorbidities have an important effect on costs, e.g. unit costs of patients with diabetes are lower than the average unit cost (Table 2).
Mean annual cost per patient shows important variations across regions. On average, the cost per patient is higher in the southern and northern regions than in the central region (Table 2), despite the proportion of dropped out patients (who are, on average, costlier) being very similar across all areas.
Ageing produces a rise in costs (Fig. 2), however the growing trend shows an inflection point in the 70–74 age group. Costs are higher for males (€2779), but the difference in the female population is negligible (+3%).
Pharmaceuticals, hospitalisation, and outpatients services account for 71.5, 18.8 and 9.7% of total healthcare expenditure, respectively (Table 3); hence, pharmaceuticals are the most important component of healthcare expenditure. Erythropoietins are the major component of drug expenditure (accounting for 36.4% of total drug expenditure), followed by antihypertensive drugs (18.7%) and drugs for CKD [12.1%, including vitamin D (9.6%) and chelating agents (2.5%)].
The contribution of hospitalisation to total costs is higher for patients who dropped out than for patients who completed the follow-up. More advanced is the starting CKD stage, the higher the proportion of inpatient costs (from 12.1% of total costs in stage 1 to 24% in stage 5) (Table 3).
Table 4 illustrates the results of the multiple regression model. The dependent variables are per capita total and annual drug expenditure. Explanatory variables are listed in column 1, while columns 2 and 3 include the reference variable and other variables, respectively. In columns 4 and 5, correlation coefficients for total healthcare and drug costs are reported. The explanatory power of the model is not particularly high (R-squared ranges from 0.136 to 0.165). However, coefficients are significant for some variables; a higher starting CKD stage, together with the presence of three comorbidities, dropout status, and belonging to the southern region are significantly correlated with both per capita healthcare and drug expenditure. These variables seem to explain more healthcare and drug expenditure than other variables (i.e. age, sex, and the presence of proteinuria at recruitment).