In our study we compared two hospitals that differed in the attainment of the PTH treatment goal and explored possible explanations. We showed that the apparent differences were explained by incomparable patient populations and laboratory techniques, and the use of the abbreviated MDRD formula to estimate GFR. After correction for kidney transplant status and with the use of the MDRD-6 formula with calibrated creatinine to estimate GFR, treatment center no longer influenced the attainment of the PTH treatment goal.
We evaluated potential differences in treatment which might explain any difference in attainment of the PTH target. We focused on characteristics of the treatment given to patients in the year before the start of the MASTERPLAN study. Plantinga et al. have shown that more frequent patient-physician contacts in patients with end-stage renal disease are positively associated with the achievement of clinical performance targets, including targets of bone and mineral disorder
. One would therefore expect that patients in center A, the ‘worst performing’ hospital, visited their physician less often and had less laboratory tests done. Table
2 shows that the opposite was true. Admittedly, the number of visits and laboratory tests may be dictated by patient morbidity and disease history and not necessarily reflect QoC, e.g. kidney transplant recipients in general need more frequent control. In addition, there were no differences in the way both centers handled the patients with PTH values exceeding recommended levels.
2 also shows that relatively few PTH tests were performed, while PTH is one of the most deviant laboratory values in CKD patients. In both centers in only a small number of patients with PTH exceeding recommended levels, treatment was adjusted or started. Thus, although guidelines give attention to treatment of CKD- bone and mineral disease, and although treatment targets are well defined, physicians are insufficiently aware of the importance of adequate treatment of hyperparathyroidism: our data point to therapeutic inertia towards the PTH treatment target.
In univariate analysis several factors were potential determinants of PTH. Multivariate analyses showed that many of these factors did not independently predict PTH levels. Renal function on the basis of calibrated creatinine values and kidney transplant status are the most important determinants of attainment of the PTH treatment target.
As mentioned before, GFR is a well known determinant of PTH and increases in PTH occur early in the course of renal insufficiency. We observed an inverse relationship between PTH and eGFR (Table
3). Although eGFR proved to be an important, independent predictor of PTH, in the initial analysis the differences in PTH levels between centers could not be explained by differences in eGFR. However, all formulas for estimating GFR have limitations. We showed that MDRD-4 is invalid in patients with proteinuria, where MDRD-6 proved better
. MDRD-6 superiority was also shown in kidney transplant recipients
. Since MDRD formulas critically depend on the measurement of serum creatinine, differences between serum creatinine assays affect their performance. Therefore, recent guidelines suggest to use calcibrated serum creatinine values
Our data clearly show that the use of the MDRD-6 formula reduced the center effect. Moreover, when using the MDRD-6 formula and calibrated serum creatinine values, there were no longer significant differences in attainment of PTH treatment targets between the centers.
From these findings we conclude that it is not always valid to use the abbreviated MDRD formula instead of the six-variable MDRD formula, especially in analyses in which GFR plays a central role. Moreover, comparable creatinine assays must always be used. Physicians should be aware of the limitations of formulas for estimating GFR, especially of the MDRD-4 formula, since this formula is extensively used (in The Netherlands and also in many other countries).
Admittedly, although the results may be explained by the better performance of the MDRD-6 formula as measure of real GFR, we cannot exclude that other factors are involved. The MDRD-6 formula incorporates serum albumin concentration. It is known that there is an inverse relationship between plasma calcidiol levels and magnitude of proteinuria, and thus hypoalbuminemia, because of loss of vitamin D metabolites and vitamin D binding protein in the urine in patients with proteinuria
[34, 35]. Low plasma calcidiol levels are related to higher PTH concentrations
It is well known that hyperparathyroidism often persists for many years after kidney transplantation
[40, 41]. Treatment with vitamin D compounds is complicated, since hyperparathyroidism in post-transplant patients is usually associated with hypercalcemia
. Consequently, mean PTH level was higher in kidney transplant recipients and only a smaller number of kidney transplant patients achieved the PTH treatment goal. In center A more kidney transplant recipients were treated than in center B.
Insight and transparency in QoC is becoming more and more important. Various indicators are used to assess hospital performance, for example attainment of treatment targets, as in our study. Benchmarking has become a way to compare and judge treatment centers. Ranking hospitals on the basis of performance indicators is supposed to give health care professionals, insurance companies as well as (associations of) patients insight in QoC. The results of our study question the reliability of these rankings and other hospital performance comparisons.
There are several difficulties associated with hospital performance comparisons: definitions are not always the same
[16, 42], laboratory assays vary
, data quality is variable between hospitals and even within one hospital
[16, 42, 43]. Another problem is patient case mix
[16, 18, 42, 43]. Patient age, race, severity of illness, and comorbidity all influence the outcome of care. Retrospective risk adjustment can only partly adjust for all these factors. There will always be additional residual confounding
[42, 43]. Moreover, random variation has to be taken into account
[16, 42–45]. Therefore, when describing results of hospital performance comparisons, confidence intervals should be provided to give insight into the influence of random variation
[16, 44]. The practice of summarizing several performance indicators in one composite score adds to the unreliability of performance measures since small differences in methods of constructing the composite score can have substantial impact on the results
. Finally, whether performance indicators provide a true reflection of QoC is questionable. Performance indicators represent mainly technical aspects, while the humane side of health care and traditional components of caring, essential when talking about QoC, are ignored
A major limitation of our study is that we compared only two hospitals. The extent to which the findings can be generalized beyond the centers and cases studied, is unknown. Another limitation is the lack of plasma calcidiol levels. Since plasma calcidiol and PTH are inversely related
[36–39], differences in calcidiol levels can have important consequences for PTH concentrations. Other unknown, and possibly influencing, factors are patient compliance, FGF-23 levels, and dietary intake of calcium, phosphate, protein and vitamin D, including over-the-counter vitamin pills.