Polypharmacy and medication use in patients with chronic kidney disease with and without kidney replacement therapy compared to matched controls.

INTRODUCTION

Polypharmacy (PP), defined as the concomitant use of medications by one individual, is a frequent phenomenon in clinical practice [ 1 , 2 ]. Older age and multimorbidity are associated with the growing PP prevalence [ 2–4 ]. Chronic kidney disease (CKD) patients often have a large burden of comorbidities and commonly require a multitude of medications to prevent further progression of CKD, to treat its complications and to treat comorbidities [ 5 ]. This makes PP a part of their life [ 6–8 ]. PP puts patients at risk of medication-related problems, such as drug–drug interactions, suboptimal therapeutic response, a higher risk of adverse drug events and decreased medication adherence [ 5 , 9 ]. Additionally, PP is associated with poorer quality of life, increased healthcare utilization with higher healthcare costs and a higher risk of morbidity and mortality [ 2 , 10 , 11 ]. Whether the poor outcomes associated with PP are merely a reflection of a person’s poor health remain unclear. Nevertheless, findings from previously published papers suggest an association between PP and mortality, even after adjustment for measured confounders such as comorbidities [ 12 ].

The prevalence of PP varies across countries and stages of CKD [ 6–8 , 10 , 13–17 ]. Current studies mostly report on elderly patients, only a few studies have used nationwide data and most studies lack a comparison with the general population [ 6 , 7 , 15 ]. This study aims to examine PP in patients with CKD Stage G4/G5 and patients on kidney replacement therapy (KRT) compared with matched general population controls of similar age, sex and socio-economic status (SES), while making use of a national health insurance database encompassing the complete known Dutch kidney disease population. Furthermore, we aim to determine risk factors for PP and commonly dispensed medications.

MATERIALS AND METHODS

Vektis insurance claims database

We used the Vektis database, which includes virtually all Dutch citizens [ 18 ]. Vektis contains reimbursement data on all medical procedures covered by the Health Insurance Act and demographic data such as sex, year of birth, area of residence, SES ( Appendix 1 ) and date of death [ 19 ].

All hospital procedures in The Netherlands are reimbursed via physician claims called Diagnosis–Treatment Combinations (DBCs) [ 20 ]. Vektis also includes pharmacy dispensing data on anatomical therapeutic chemical code level, the defined daily dose (DDD) and the quantity of supplied medication per year. A DDD is a technical unit that reflects the assumed average maintenance dose per day for a medication used for its main indication [ 21 ]. The annual quantity supplied for a specific medication is a product of the DDD and the number of days a medication was dispensed. Information on over-the-counter medications and medications administered during a hospital admission or dialysis treatment are missing, since the costs for the latter are covered by the hospital DBC. Since health claims databases lack clinical data, we used proxies [e.g. pharmaceutical cost groups (PCGs)], to assess the prevalence and number of chronic conditions ( Appendix 1 ) [ 22 , 23 ]. Hospitalization, intensive care unit (ICU) admission and emergency room (ER) visits were identified by specific healthcare operation codes, an element of the DBC code ( Appendix 1 ).

Study population

We selected adults (i.e. ≥20 years) with CKD Stage G4/G5 or on KRT using 2017 healthcare claims data. Patients were divided into three patient groups: CKD Stage G4/G5 [estimated glomerular filtration rate (eGFR) <30 mL/min/1.73 m 2 ] without KRT, dialysis patients and kidney transplant patients. Patients were excluded if they switched between groups in 2017 (i.e. from CKD Stage G4/G5 to KRT and vice versa or between KRT modalities), if they died in 2017 or if matching was impossible ( Figure 1 ).

CKD Stage G4/G5 without KRT

We selected patients with a CKD Stage G4/G5 health claim on 1 January 2017. Since primary care does not have ‘disease-specific’ claims comparable to DBCs, we could not identify patients solely treated in primary care.

Dialysis

Patients with a health claim for dialysis on 1 January 2017 were selected regardless of dialysis modality.

Kidney transplantation

Patients with a health claim for kidney transplantation on 1 January 2017 were selected.

Control groups

Two controls per patient, matched for age, sex and SES (per quartile) were selected, provided they had no CKD-related healthcare claim.

PP

Medications with a cumulative annual DDD ≥15 (except for antibiotic treatment) and medications with a cumulative annual DDD ≥180 were selected. The first group (DDD ≥15) was further indicated as ‘all medication use’, to prevent inclusion of medication dispensed for a very short period, and the second cut-off (DDD ≥180) as ‘chronic medication use’.

We defined PP at three levels: concurrent use of ≥5 medications (PP), ≥10 medications [excessive PP (EPP)] and ≥15 medications [hyper PP (HPP)]. For combination medications, the individual substances could not be extracted and therefore were counted as one.

Statistical analysis

To describe baseline characteristics we used means and standard deviations (SDs) for continuous variables and frequency distributions with percentages for categorical variables. To compare baseline characteristics between patients and controls we used the chi-squared test for categorical variables and the Mann–Whitney U-test for non-normally distributed continuous variables. We calculated the PP, EPP and HPP prevalences in all patient (sub)groups and controls and expressed them as percentages. These analyses were repeated in a sensitivity analysis, including all patients who died in 2017. Ratios were calculated by dividing the PP prevalence of patients by the respective prevalence in controls. Univariate and multivariate logistic regressions were used to analyse the association between the independent variables [e.g. age, sex and diabetes mellitus (DM)] and the outcome (i.e. EPP based on chronic medication use). The EPP prevalence was low (i.e. ≤15%) and therefore the rare disease assumption for logistic regression was met [ 24 ]. For the identification of confounders, we took the criteria for confounding into account [ 25 ]. Associations were expressed as odd ratios (ORs) with 95% confidence intervals (CIs). We considered a P-value <0.05 as statistically significant. Analyses were performed in SAS version 9.4 (SAS Institute, Cary, NC, USA).

RESULTS

Baseline characteristics

We included 27 573 individuals: 14 905 CKD Stage G4/G5 without KRT, 3872 dialysis and 8796 transplant patients, with mean ages of 75.6, 70.8 and 56.5 years, respectively ( Table 1 ).

Chronic comorbidity conditions were 2.9 times more prevalent in CKD Stage G4/G5 patients than in controls (1.92 versus 0.68), 3.0 times higher in dialysis patients (1.86 versus 0.61) and 4.4 times higher in transplant patients (1.46 versus 0.33). In all patient groups, the prevalence of DM, macrovascular disease and hypertension was significantly higher than in controls.

Number of dispensed medications

All medication use

The median number of dispensed medications was 10 for CKD Stage G4/G5 patients, 12 for dialysis patients and 11 for transplant patients compared with 1, 1 and 0 in controls, respectively ( Figure 2 ).

Chronic medication use

The median number of dispensed medications was six in all patient groups, compared with zero in controls ( Figure 3 ).

PP

Figure 4 presents the prevalence and ratio of PP in patients versus controls for ‘all medication use’ (left panel) and ‘chronic medication use’ (right panel). The results of the sensitivity analyses were consistent with the results of the main analyses ( Appendix 2 ).

Overall

All medication use

The PP, EPP and HPP prevalences were 87.4, 56.6 and 22.8%, respectively, in patients with CKD Stage G4/G5; 93.4, 69.3 and 31.5%, respectively, in dialysis patients; and 94.8, 60.0 and 21.5%, respectively, in transplant patients (Figure 4). For all comparisons, the PP, EPP and HPP prevalences were much higher in patients than in controls, with ratios ranging from 2.6 (PP in CKD patients versus controls) to 23.9 (EPP in transplant patients versus controls).

Chronic medication use

Overall, PP based on chronic medication use was less common than PP based on all medication use ( Figure 4 ). The PP, EPP and HPP prevalences were 66.1, 13.3 and 0.9%, respectively, in CKD Stage G4/G5 patients; 70.0, 15.1 and 1.2%, respectively, in dialysis patients; and 75.0, 14.9 and 1.0%, respectively, in transplant patients. Ratios ranged from 3.7 (PP in CKD patients) to 25.8 (EPP in transplant patients).

Patient subgroups

Tables 2 and 3 show the prevalence and ratio of PP in patients versus controls for different subgroups and for ‘all’ and ‘chronic medication use’. Since the PP prevalence for ‘all medication use’ was very high and the HPP prevalence for ‘chronic medication use’ was very low, these results are not shown.

All medication use

In CKD Stage G4/G5 and in transplant patients, the EPP and HPP prevalences were highest in patients ≥75 years of age (CKD G4/G5: 60.0 and 24.4%; transplantation: 77.4 and 34.2%). EPP was 42.0 times more common in young CKD patients (ages 20–44 years) than in controls, and this ratio declined with age to 3.8 in patients ≥75 years ( Tables 2 ). PP was more common in both patients and controls with chronic conditions, such as diabetes or macrovascular disease, with EP prevalence ranging from 78.1 to 89.8% in patient groups and 24.6 to 47.5% in controls. The highest PP prevalence (EPP 90.8%) was found in transplant patients with coronary artery disease.

Chronic medication use

PP was most common in CKD patients (69.4%) and dialysis patients (73.5%) ages 65–74 years and in transplant patients (85.0%) ≥75 years of age. Ratios between patient and control groups decreased with increasing age. The prevalence of PP was high in patients with chronic conditions in all patient groups ( Table 3 ).

Risk factors for PP

Table 4 presents the unadjusted and adjusted association between patient demographics and disease-related variables and EPP (≥10 medications, ‘chronic medication use’). Below we discuss the fully adjusted models if adjustment for potential confounders was possible.

CKD Stage G4/G5 without KRT

Patients ages 65–74 years [OR 1.57 (95% CI 1.33–1.85)] and ≥75 years [OR 1.24 (95% CI 1.06–1.44)] had a higher EPP risk compared with patients ages 20–64 years. In addition, an SES score in the lowest two quartiles compared with an SES score in the highest quartile [OR 1.34 (95% CI 1.17–1.55) versus OR 1.23 (95% CI 1.07–1.43)], diabetes [OR 4.98 (95% CI 4.51–5.54)] or vascular disease [OR 2.01 (95% CI 2.12–2.62)], as well as hospitalization [OR 1.35 (95% CI 1.17–1.55)] and an ER visit [OR 1.69 (95% CI 1.53–1.88)] were significantly associated with PP.

Dialysis

Patients ≥75 years of age had a lower risk of EPP [OR 0.74 (95% CI 0.59–0.91)] compared with patients ages 20–64 years. The most pronounced risk factors for EPP in dialysis patients were diabetes [OR 3.69 (95% CI 3.08–4.43)] and vascular disease [OR 2.08 (95% CI 1.72–2.51)].

Kidney transplantation

Patients ages 65–74 years [OR 3.69 (95% CI 2.89–4.71)] and ≥75 years [OR 5.88 (95% CI 4.60–7.51)] had a higher EPP risk compared with patients ages 20–64 years. In addition, being male [OR 1.19 (95% CI 1.05–1.34)], having an SES score in the lowest two quartiles compared with an SES score in the highest quartile [OR 1.34 (95% CI 1.13–1.59) versus OR 1.29 (95% CI 1.09–1.54)], diabetes [OR 5.59 (95% CI 4.91–6.36)] or vascular disease [OR 2.51 (95% CI 2.14–2.96)], hospitalization [OR 1.29 (95% CI 1.09–1.52)] and an ER visit [OR 1.76 (95% CI 1.54–2.00)] were significantly associated with EPP.

Dispensed medication classes

Table 5 shows the most commonly dispensed chronic medication. Proton pump inhibitors (PPIs) were among the most commonly dispensed medications in patients, with ≥50% of patients using a PPI versus 8–19% of controls. Also, statins were commonly dispensed (53, 51 and 40% in CKD Stage G4/G5, transplant and dialysis patients, respectively). Dispensed medication classes for all medication use are shown in Appendix 3 . Of note, 3–12% of CKD patients with DM do not use antidiabetic medication, whereas 17–19% of controls with DM are diet-controlled ( Appendix 3 , Table A1). Furthermore, 63–75% of CKD patients with DM chronically use antidiabetic medication compared with 61–65% of controls ( Table 5 ).

DISCUSSION

This study using Dutch health claims data demonstrates that PP is highly prevalent in CKD Stage G4/G5 patients and patients with KRT compared with the general population. Since multimorbidity is one of the driving factors of PP, we must note that chronic comorbid conditions were three to four times more prevalent in patients than in controls. In our study, PP prevalence based on ‘all medication use’ ranged from 87% in CKD Stage G4/G5 to 94–95% in dialysis and transplant patients. The prevalence was lower for chronic medication use. Older age was an important risk factor for PP in CKD Stage G4/G5 and transplant patients, whereas dialysis patients ≥75 years of age had a lower risk of PP compared with younger counterparts. For all patients, additional risk factors were lower SES, DM, vascular disease, hospitalization and an ER visit during the year. The PP prevalence ratio between patients and controls declined with age. The most commonly dispensed medications were PPIs and statins, with more than half of patients using these medications.

Strengths and limitations

The main strength of this article is the use of a health claims database with almost complete national coverage of Stage G4/G5 CKD patients, by which we could study CKD Stage G4/G5 and KRT patients in the same cohort and compare them with the general population. Pharmacy dispensing data were complete and contained all medication dispensed by the pharmacy. This in contrast to other studies that used data from patient questionnaires, which heavily relies on patient memory. Another strength of pharmacy dispensing data is that they only include prescribed medication that was actually dispensed and do not cover prescribed medications that were never collected at the pharmacy. Although information on medication adherence is often missing in studies describing medication use, the regular dispensing of medication in a health claims database is an indirect yet strong indication that the medication was routinely taken.

We must consider several limitations. First, although the identification of dialysis and transplant patients is accurate using health claims data [ 26 ], we were unable to identify patients with CKD treated in primary care, being mostly elderly patients [ 27 ]. Furthermore, data on medication adherence are missing. In addition, we were unable to identify medication given during dialysis sessions. Therefore the PP levels of dialysis patients reported in this study are likely an underestimation of their actual medication burden. Finally, the estimation of chronic conditions in our study was based on proxies that are vulnerable to inaccuracy.

Prevalence of PP

The comparison of the prevalence of PP with other studies is challenging due to the substantial differences in patient selection, definition of PP and data collection. Almost all previously performed studies collected cross-sectional medication data via patient reports or medical charts. Our study is unique in that we used pharmacy dispensing data, which enabled us to monitor all dispensed medication. The availability of the annual quantity of supplied medications makes it possible to differentiate between all and chronic medication use.

The considerably higher PP prevalence based on all medication use compared with chronic medication use suggests that patients often receive short-term medication or experience medication changes. Although PP prevalence based on chronic medication use better reflects the structural medication burden, this type of medication use is not reported in other studies. Therefore we can only discuss our findings on the PP prevalence in the perspective of other studies on all medication use.

CKD Stage G4/G5 without KRT

Current literature describes PP prevalence in different stages of CKD, using different definitions of PP and mainly in elderly patients. Two studies describe PP prevalence in CKD Stage G4/G5 patients. Of these, Schmidt et al. [ 6 ] reported a PP prevalence of 92% (eGFR <30mL/min/1.73 m 2 ). Hayward et al. [ 15 ] describe prevalences of 91% (≥5 medications) and 43% (≥10 medications) in a group of elderly (age >65 years) patients (eGFR <20 mL/min/1.73 m 2 ) of different European countries. Within the subset of Dutch patients in this study, a prevalence of 91% (≥5 medications) and 43% (≥10 medications) was described. All results are comparable to our findings. Lower PP prevalence was found in patients with CKD Stages G1–G3 [ 6–8 ].

Dialysis

It is well known that dialysis patients have a high medication burden [ 13 , 28 , 29 ]. A pooled analysis reported that dialysis patients use 12 different medications [ 10 , 29 ]. We report a median of 12 medications. A study from Saudi Arabia with 95 haemodialysis patients reported a 98% PP prevalence (>5 medications) [ 16 ], which is comparable to our PP prevalence. A Canadian study reported that 93.1% of elderly haemodialysis patients (age ≥65 years) used five or more medications [ 10 ]. No previous studies have reported on EP and HP prevalence and we are the first study in a much larger cohort of dialysis patients of all ages.

Kidney transplantation

A high pill burden is also described in transplant patients, ranging from 7 to 32 pills per day, depending on the time period after transplantation [ 30–32 ]. An Argentinean study described a mean of 7.8 different medications, while we describe a median of 10 different medications [ 33 ]. Only one Polish study reported PP prevalence in a much smaller group of 136 transplant patients as 56% (5–9 medications) and 17% (≥10 medications) [ 17 ]. We demonstrated a considerably higher PP and EP prevalence in our larger cohort of transplant patients.

Comparison with the general population

To our knowledge, this is the first study comparing the PP prevalence of CKD Stage G4/G5 patients and KRT patients with a matched control group from the general population. We demonstrate that PP prevalence is already substantially higher in young patients compared with controls, probably reflecting the high number of comorbidities in CKD patients already at a young age. The ratio of PP between patients and controls decreases with increasing age, because medication use increases more with age in the general population than it does in patients [ 34 ].

Risk factors for PP

We confirm a positive association between PP and older age in CKD Stage G4/G5 and transplant patients [ 6 , 8 , 17 , 35 ]. The inverse association between PP and age ≥75 years in dialysis patients may suggest some reluctance to prescribe medication in the elderly dialysis patient with limited life expectancy and being at high risk for medication-related problems. We confirm that the presence of chronic conditions like DM and cardiovascular disease are risk factors for PP in all patients [ 6 , 10 , 16 , 36 ].

Next, we described a positive association between low SES and PP for CKD Stage G4/G5 and transplant patients, in line with other studies [ 6 , 8 ]. A possible explanation is that individuals with a low SES often have low health literacy and are more vulnerable to comorbid illness. Lastly, we are the first to demonstrate a positive association between PP and hospitalization or an ER visit. We hypothesize that patients with an indication for an ER visit or hospital admission likely have severe comorbid conditions or complications of their CKD for which they need additional medication prescriptions. Moreover, PP itself may be associated with hospitalization in the elderly population [ 37 , 38 ], although this was not confirmed elsewhere [ 39 ].

Medication dispensing

The increased cardiovascular risk of CKD patients is reflected in the high number of medications to prevent or treat cardiovascular conditions. Recent guidelines recommend statin prescription to CKD Stage G4/G5 patients [ 40 ]. Although (almost) all CKD Stage G4/G5 patients would be expected to fulfil the criteria for statin prescription, only half of the patients in our study used statins. Conversely, several studies question the benefit of statin therapy for dialysis patients [ 41–43 ]. Guidelines suggest that statins should not be routinely ‘initiated’, though they should be continued when patients already use statins when initiating dialysis treatment [ 44 ]. We suggest a critical evaluation of statin treatment in dialysis patients to reduce some of the medication burden. This also may be the case for PPIs [ 45 ]. More than 50% of CKD Stage G4/G5 and transplant patients, and even >65% of dialysis patients, used a PPI in our study. Previous studies reported PPI use of 30, 50 and 52% in haemodialysis patients and 33, 49 and 62% in CKD Stage G4/G5 patients, respectively [ 10 , 15 , 36 ]. The literature reports that the indication for PPI use in dialysis patients was unknown >25% of the time [ 46 ]. Since the long-term use of PPIs can have negative consequences, deprescribing of PPIs should be considered [ 47 ].

CONCLUSION

Our study demonstrates that patients with CKD Stage G4/G5 and patients on KRT have a very high medication burden, far beyond that of individuals from the general population. Important PP risk factors are age, SES, DM, vascular disease, hospitalization and an ER visit.

Medication treatment of CKD patients is a challenging balance between the benefits of pharmacotherapy for the treatment of kidney disease and comorbidities and the disadvantages of potentially inappropriate prescribing or adverse drug interaction [ 48 ]. Although physicians often check whether the prescribed medication is appropriate in their patient, it is not easy to minimize the medication burden. As directed by the Hippocratic Oath, physicians strive for optimal treatment of their patients, while avoiding those twin traps of overtreatment and therapeutic nihilism. Undertreatment has been repeatedly associated with unfavourable outcomes in dialysis patients [ 49 ]. Despite the fact that therapeutic nihilism should be avoided at all times, we propose that a critical approach to the prescription of specific medications like PPIs in all CKD patients and statins in the dialysis population could be a first step towards more appropriate medication use. Finding a proper balance between potentially beneficial medication and needless use of medications with adverse effects will remain a challenge.

FUNDING

This work is financed by a grant from the Dutch Kidney Foundation.

CONFLICT OF INTEREST STATEMENT

None declared.

DATA AVAILABILITY STATEMENT

The Vektis database used for this study can only be accessed by contacting Vektis (see www.vektis.nl ).

REFERENCES

1 Fincke BG , Snyder K , Cantillon C et al Three complementary definitions of polypharmacy: methods, application and comparison of findings in a large prescription database . Pharmacoepidemiol Drug Saf 2005 ; 14 : 121 – 128 15386712 10.1002/pds.966

2 Payne RA. The epidemiology of polypharmacy . Clin Med (Lond) 2016 ; 16 : 465 – 469 27697812 10.7861/clinmedicine.16-5-465 PMC6297306

3 Morin L , Johnell K , Laroche ML et al The epidemiology of polypharmacy in older adults: register-based prospective cohort study . Clin Epidemiol 2018 ; 10 : 289 – 298 29559811 10.2147/CLEP.S153458 PMC5856059

4 Fano V. Estimating the prevalence and the determinants of polypharmacy using data from a health administrative database: a comparison of results obtained employing different algorithms . Adv Pharmacoepidemiol Drug Saf 2014 ; 3 : 151

5 Mason NA , Bakus JL. Strategies for reducing polypharmacy and other medication-related problems in chronic kidney disease . Semin Dial 2010 ; 23 : 55 – 61 19747171 10.1111/j.1525-139X.2009.00629.x

6 Schmidt IM , Hübner S , Nadal J et al Patterns of medication use and the burden of polypharmacy in patients with chronic kidney disease: the German Chronic Kidney Disease study . Clin Kidney J 2019 ; 12 : 663 – 672 31584562 10.1093/ckj/sfz046 PMC6768303

7 Laville SM , Metzger M , Stengel B et al Evaluation of the adequacy of drug prescriptions in patients with chronic kidney disease: results from the CKD-REIN cohort . Br J Clin Pharmacol 2018 ; 84 : 2811 – 2823 30110711 10.1111/bcp.13738 PMC6255993

8 Fraser SDS , Roderick PJ , May CR et al The burden of comorbidity in people with chronic kidney disease stage 3: a cohort study . BMC Nephrol 2015 ; 16 : 193 26620131 10.1186/s12882-015-0189-z PMC4666158

9 Mason NA. Polypharmacy and medication-related complications in the chronic kidney disease patient . Curr Opin Nephrol Hypertens 2011 ; 20 : 492 – 497 21788893 10.1097/MNH.0b013e328349c261

10 Battistella M , Jandoc R , Ng JY et al A province-wide, cross-sectional study of demographics and medication use of patients in hemodialysis units across Ontario . Can J Kidney Health Dis 2018 ; 5 : 2054358118760832 29568537 10.1177/2054358118760832 PMC5858734

11 Park HY , Ryu HN , Shim MK et al Prescribed drugs and polypharmacy in healthcare service users in South Korea: an analysis based on National Health Insurance Claims data . Int J Clin Pharmacol Ther 2016 ; 54 : 369 – 377 27007996 10.5414/CP202484

12 Leelakanok N , Holcombe AL , Lund BC et al Association between polypharmacy and death: a systematic review and meta-analysis . J Am Pharm Assoc 2017 ; 57 : 729 – 738.e10 10.1016/j.japh.2017.06.002 28784299

13 Parker K , Nikam M , Jayanti A , Mitra S. Medication burden in CKD-5D: impact of dialysis modality and setting . Clin Kidney J 2014 ; 7 : 557 – 561 25859372 10.1093/ckj/sfu091 PMC4389130

14 St Peter WL. Management of polypharmacy in dialysis patients . Semin Dial 2015 ; 28 : 427 – 432 25864928 10.1111/sdi.12377

15 Hayward S , Hole B , Denholm R et al International prescribing patterns and polypharmacy in older people with advanced chronic kidney disease: results from the European Quality study . Nephrol Dial Transplant 2020 ; 36 : 503 – 511 10.1093/ndt/gfaa064 32543669

16 Alshamrani M , Almalki A , Qureshi M et al Polypharmacy and medication-related problems in hemodialysis patients: a call for deprescribing . Pharmacy (Basel) 2018 ; 6 : 76 10.3390/pharmacy6030076 PMC6163628 30046021

17 Woźniak I , Kolonko A , Chudek J et al Influence of polypharmacy on the quality of life in stable kidney transplant recipients . Transplant Proc 2018 ; 50 : 1896 – 1899 30056924 10.1016/j.transproceed.2018.02.128

18 Vektis. Vektis - Inzichten op maat. www.vektis.nl (3 March 2020, date last accessed)

19 de Boo A. Vektis - information center for health care services . TSG 2011 ; 89 : 358 – 359

20 Westerdijk M , Zuurbier J , Ludwig M , Prins S. Defining care products to finance health care in the Netherlands . Eur J Health Econ 2012 ; 13 : 203 – 221 21350859 10.1007/s10198-011-0302-6 PMC3286594

21 World Health Organization . Defined Daily Dose. Definition and General Considerations . Geneva: World Health Organization. 2021. https://www.who.int/tools/atc-ddd-toolkit/about-ddd

22 Lamers LM. Pharmacy costs groups: a risk-adjuster for capitation payments based on the use of prescribed drugs . Med Care 1999 ; 37 : 824 – 830 10448725 10.1097/00005650-199908000-00012

23 Lamers LM , van Vliet RCJA. The Pharmacy-based Cost Group model: validating and adjusting the classification of medications for chronic conditions to the Dutch situation . Health Policy 2004 ; 68 : 113 – 121 15033558 10.1016/j.healthpol.2003.09.001

24 Greenland S , Thomas DC. On the need for the rare disease assumption in case-control studies . Am J Epidemiol 1982 ; 116 : 547 – 553 7124721 10.1093/oxfordjournals.aje.a113439

25 Jager KJ , Zoccali C , Macleod A et al Confounding: what it is and how to deal with it . Kidney Int 2008 ; 73 : 256 – 260 17978811 10.1038/sj.ki.5002650

26 Mohnen SM , van Oosten MJM , Los J et al Healthcare costs of patients on different renal replacement modalities – analysis of Dutch health insurance claims data . PLoS One 2019 ; 14 : e0220800 31415578 10.1371/journal.pone.0220800 PMC6695145

27 van Oosten MJM , Brohet RM , Logtenberg SJJ et al The validity of Dutch health claims data for identifying patients with chronic kidney disease: a hospital-based study in the Netherlands . Clin Kidney J 2021 ; 14 : 1586 – 1593 34276977 10.1093/ckj/sfaa167 PMC8280937

28 Chiu YW , Teitelbaum I , Misra M et al Pill burden, adherence, hyperphosphatemia, and quality of life in maintenance dialysis patients . Clin J Am Soc Nephrol 2009 ; 4 : 1089 – 1096 19423571 10.2215/CJN.00290109 PMC2689877

29 Manley HJ , Cannella CA , Bailie GR et al Medication-related problems in ambulatory hemodialysis patients: a pooled analysis . Am J Kidney Dis 2005 ; 46 : 669 – 680 16183422 10.1053/j.ajkd.2005.07.001

30 Hardinger KL , Hutcherson T , Preston D et al Influence of pill burden and drug cost on renal function after transplantation . Pharmacotherapy 2012 ; 32 : 427 – 432 22467382 10.1002/j.1875-9114.2012.01032.x

31 Adhikari UR , Taraphder A , Hazra A et al Pill burden does not influence compliance with oral medication in recipients of renal transplant . Indian J Pharmacol 2016 ; 48 : 21 – 25 26997717 10.4103/0253-7613.174425 PMC4778200

32 Low JK , Crawford K , Manias E et al Quantifying the medication burden of kidney transplant recipients in the first year post-transplantation . Int J Clin Pharm 2018 ; 40 : 1242 – 1249 29956133 10.1007/s11096-018-0678-9

33 Bril F , Castro V , Centurion IG et al A systematic approach to assess the burden of drug interactions in adult kidney transplant patients . Curr Drug Saf 2016 ; 11 : 156 – 163 27194037 10.2174/157488631102160429003742

34 Cadogan CA , Ryan C , Hughes CM. Appropriate polypharmacy and medicine safety: when many is not too many . Drug Saf 2016 ; 39 : 109 – 116 26692396 10.1007/s40264-015-0378-5 PMC4735229

35 Schuler J , Dückelmann C , Beindl W et al Polypharmacy and inappropriate prescribing in elderly internal-medicine patients in Austria . Wien Klin Wochenschr 2008 ; 120 : 733 – 741 19122984 10.1007/s00508-008-1089-z

36 Manley HJ , Garvin CG , Drayer DK et al Medication prescribing patterns in ambulatory haemodialysis patients: comparisons of USRDS to a large not-for-profit dialysis provider. Nephrol Dial Transplant 2004 ; 19 : 1842 – 1848 15128886 10.1093/ndt/gfh280

37 Parker K , Wong J. Is polypharmacy an increasing burden in chronic kidney disease? The German experience . Clin Kidney J 2019 ; 12 : 659 – 662 31584570 10.1093/ckj/sfz072 PMC6768297

38 Fried TR , O’Leary J , Towle V et al Health outcomes associated with polypharmacy in community-dwelling older adults: a systematic review . J Am Geriatr Soc 2014 ; 62 : 2261 – 2272 25516023 10.1111/jgs.13153 PMC4270076

39 Payne RA , Abel GA , Avery AJ et al Is polypharmacy always hazardous? A retrospective cohort analysis using linked electronic health records from primary and secondary care . Br J Clin Pharmacol 2014 ; 77 : 1073 – 1082 24428591 10.1111/bcp.12292 PMC4093932

40 NHG-Richtlijen. Chronische nierschade . https://richtlijnen.nhg.org/standaarden/chronische-nierschade

41 Wanner C , Krane V , März W et al Atorvastatin in patients with type 2 diabetes mellitus undergoing hemodialysis . N Engl J Med 2005 ; 353 : 238 – 248 16034009 10.1056/NEJMoa043545

42 Baigent C , Landray MJ , Reith C et al The effects of lowering LDL cholesterol with simvastatin plus ezetimibe in patients with chronic kidney disease (Study of Heart and Renal Protection): a randomised placebo-controlled trial . Lancet 2011 ; 377 : 2181 – 2192 21663949 10.1016/S0140-6736(11)60739-3 PMC3145073

43 Fellström BC , Jardine AG , Schmieder RE et al Rosuvastatin and cardiovascular events in patients undergoing hemodialysis . N Engl J Med 2009 ; 360 : 1395 – 1407 19332456 10.1056/NEJMoa0810177

44 Wanner C , Tonelli M , Kidney Disease: Improving Global Outcomes Lipid Guideline Development Work Group Members. KDIGO clinical practice guideline for lipid management in CKD: summary of recommendation statements and clinical approach to the patient . Kidney Int 2014 ; 85 : 1303 – 1309 24552851 10.1038/ki.2014.31

45 Desbuissons G , Mercadal L. Use of proton pump inhibitors in dialysis patients: a double-edged sword . J Nephrol 2021 ; 34 : 661 – 672 32710264 10.1007/s40620-020-00808-y

46 McIntyre C , McQuillan R , Bell C et al Targeted deprescribing in an outpatient hemodialysis unit: a quality improvement study to decrease polypharmacy . Am J Kidney Dis 2017 ; 70 : 611 – 618 28416321 10.1053/j.ajkd.2017.02.374

47 Triantafylidis LK , Hawley CE , Perry LP et al The role of deprescribing in older adults with chronic kidney disease . Drugs Aging 2018 ; 35 : 973 – 984 30284120 10.1007/s40266-018-0593-8 PMC8150926

48 MacRae CE , Mercer S , Guthrie B. Potentially inappropriate prescribing in people with chronic kidney disease: cross-sectional analysis of a large population cohort . Br J Gen Pract 2020 ; 71: e483 – e490 10.3399/BJGP.2020.0871 PMC8103925 33947664

49 Ortiz A , Covic A , Fliser D et al Epidemiology, contributors to, and clinical trials of mortality risk in chronic kidney failure . Lancet 2014 ; 383 : 1831 – 1843 24856028 10.1016/S0140-6736(14)60384-6

APPENDIX 1. VARIABLES BASED ON DATA OF THE VEKTIS DATABASE

SES

The SES was established by the Netherlands Institute for Social Research and is based on a person’s postal code [ 20 ]. The SES score is derived from the mean income in the residential area, the percentage of people with low education and low income as well as the fraction of unemployed people in the area. The national mean SES score is 0 and ranges from −8.07 to +3.06, where a lower score indicates a lower SES and a higher score indicates a higher SES.

DM

The definition of the variable DM is based on a combination of hospital claims (DBC codes), pharmaceutical claims and health claims for primary care activities.

Macrovascular disease, coronary artery disease, peripheral artery disease and cerebrovascular accident (CVA)/transient ischaemic attack (TIA)

The variable macrovascular disease is a combination of the variables coronary artery disease, peripheral artery disease and CVA/TIA. The definitions of the variables coronary artery disease (= 1), peripheral artery disease (= 2) and CVA/TIA (= 3) are based on hospital claims (DBC codes).

Malignancy

The definition of the variable malignancy is based on hospital claims (DBC codes).

Hypertension

The definition of the variable hypertension is based on a combination of hospital claims (DBC codes) and pharmaceutical claims.

Hospitalization

The definition of the variable hospitalization is based on health claims for hospital care activities that are linked to hospital claims (DBC codes). We excluded hospital care activities if the admission was related to transplantation care.

ICU admission

The definition of the variable ICU admissions is based on hospital declaration codes that are linked to hospital claims (DBC codes).

ER visits

The definition of the variable ER visits is based on hospital declaration codes that are linked to hospital claims (DBC codes).

Chronic conditions based on PCGs

Since clinical data are lacking in health claims databases, we used PCGs as a proxy to determine chronic conditions. PCGs are defined by the Zorginstituut Nederland (National Health Care Institute) and are used as a risk adjuster in the Dutch healthcare system [ 18 ]. Within this risk adjustment system, Dutch insurance companies receive an equalization contribution from the Healthcare Insurance Fund depending on the risk profile of the insured population. This risk profile is based on, among other things, age, gender, SES and the number of chronic conditions (PCGs), as these factors have been shown to increase the healthcare costs in subsequent years [ 21 ].

PCGs are based on the assumption that chronic conditions can be reliably identified by claims for specific prescribed drugs [ 18 , 19 ]. A person is assigned to a PCG if the prescribed medication for a chronic condition is more than a certain amount during a calendar year (e.g. 180 DDD, which approximates 6 months of medication use). The validity of pharmacy claims data to identify chronic conditions has been evaluated before and has been shown to provide reliable estimates of chronic disease burden when clinical data are missing [ 22–24 ].

Chronic conditions based on PCGS

A total of 37 PCGs for the risk adjustment of 2019 (based on pharmacy data of 2017) are defined in this section [ 25 ]. We excluded the PCGs for CKD and transplantation since these overlap with the main diagnosis of our study population. Appendix 1 (Tables A1– A 33 ) provides the chronic conditions used in this study derived from the PCGs, with the ATC codes and DDDs used for the classification of PCGs.

APPENDIX 2: SENSITIVITY ANALYSIS

Sensitivity analysis was performed in which the prevalence of PP was examined in case all deceased patients in 2017 were included.

APPENDIX 3