FormalPara Key Points

Several US studies showed a negative effect of high deductible healthcare plans and copayments on the adherence of medication to treat cardiovascular risk factors. Since the healthcare system in the USA is completely different from that of most other Western (European) systems, this study will focus on the effect of deductibles on the adherence of cardiovascular medication in the Netherlands.

The use of deductibles in the Dutch healthcare system was associated with a decrease in the adherence to antihypertensive and antihyperlipidemic monotherapies in primary prevention of cardiovascular events.

1 Introduction

In 2021, more than 37,000 (22% of all deaths in 2021 in the Netherlands) people in the Netherlands died from cardiovascular disease, which is the second leading cause of death after cancer [1]. The risk of cardiovascular diseases can be decreased by using antihypertensive and antihyperlipidemia medication in patients at risk. As shown by different meta-analyses and systematic reviews of clinical and randomized trials, these medications can substantially reduce the risk of cardiovascular events (CVEs) in both patients that already suffered from a cardiovascular event (secondary prevention) as well as those without CVE (primary prevention) [2,3,4,5]. In earlier research we found that adherence to antihyperlipidemia medication is an important modifiable factor for reducing the risk of cardiovascular events in primary prevention [6]. Besides that, adherence is lower in primary preventive patients than in secondary preventive patients [7, 8]. In addition, for antihypertensive medication, Xu et al. [9] showed that high adherence is dose-dependently associated with a low risk of CVEs.

Lewey et al. [10] showed that people who are insured with a high deductible healthcare plan have reduced adherence to medication to treat cardiovascular risk factors, regardless of socioeconomic status or race/ethnicity. Full coverage of cardiovascular medication compared to cost-sharing healthcare plans improves the adherence of these types of medication [11, 12]. A meta-analysis of Lemstra et al. [13] showed that one of the drivers of non-adherence in patients on statin therapy are copayments; using copayments increases the risk of being non-adherent by 28% compared with not using copayments. Similar associations between reduced adherence and using copayments have been found for antihypertensive drugs [14].

Several studies hence showed that there is an effect of different types of cost-sharing plans in the US and Canada on adherence [13,14,15]. As far as we know, no studies have been conducted to determine the effects of healthcare cost-sharing plans on the adherence of cardiovascular medications in Europe. Since the USA, and to a lesser extent Canada, has a different healthcare system than most other Western countries [16], a study of the effects of healthcare cost-sharing plans in European countries is needed. Therefore, we examined the association between having to pay out of pocket due to deductibles and the adherence to antihypertensives and antihyperlipidemia medication in the Netherlands.

2 Methods

2.1 Regulatory Framework

Since the introduction of the Zorgverzekeringswet (Health Insurance Act; Zvw) in 2006, every Dutch citizen is obliged to have health insurance offered by one of the Dutch health insurers. The Zvw creates a framework for regulated competition in the Dutch healthcare system based on the solidarity principle [17, 18]. In 2008 the Dutch government introduced the mandatory deductible for everyone that has health insurance, instead of a no-claim system [19]. This mandatory deductible not only has the goal of moving healthcare financing from the collective to private expenditures, but also tries to encourage people to review their need of medical help [20]. This mandatory deductible is supposed to increase every year following inflation and increasing healthcare expenditures, but has been fixed at €385 since 2016 [21, 22]. Except for general practitioner (GP), obstetrician and maternity care, almost all other care is subject to the mandatory deductible [23]. Most health insurance companies provide people with the option to choose for an extra voluntary deductible up to 500 euros (in steps of 100 euros) in exchange for a discount on their premium. Since 2009, health insurers are allowed to exclude certain care covered by both the mandatory as well as the voluntary deductible to increase the efficiency incentive [19, 23]. Some health insurance companies allow people to pay their mandatory deductible in installments. This is mostly used by people who know they will reach their deductible limit beforehand or early in a year [24]. When the deductible is paid in installments, but the deductible limit in a year is not reached, the amount that is not used will be reimbursed. This is a type of no-claim system. Finally, for people with a low income it is (in most municipalities) possible to get a collective health insurance through their municipality. This municipality in turn can choose to reinsure the deductible of the people in this collective, which means that those people do not have to pay their deductibles themselves. This construction is called a reinsured deductible.

Since every Dutch citizen is obliged to have health insurance and almost all care is covered by the basic health insurance, every health insurer has almost complete records of the type of care received while insured at this health insurer. To increase transparency for patients, healthcare providers and health insurers use the Dutch diagnose–behandel–combinatie (diagnosis–treatment–combination; DBC) system to manage declarations [25]. Unfortunately, there is no mapping from this system to international coding systems, such as International Classification of Diseases tenth revision (ICD-10).

2.2 Study Design and Data Source

We conducted a retrospective population-based inception cohort study using the claims database of Menzis Zorgverzekeraar N.V. (Menzis Health Insurer). This database contains health insurance claims from over two million people over multiple years and covers approximately 12% of the Dutch population [26]. The database is complete regarding pharmaceutical prescriptions, except for over the counter medication during the time someone is insured at Menzis Zorgverzekeraar N.V. Since this is a retrospective study of deidentified administrative data with a large number of participants, informed consent was waived [27].

2.3 Study Population

The study population consists of adult patients aged 19 years or older insured at a large health insurer in the Netherlands, who started antihyperlipidemic or antihypertensive monotherapy for the primary prevention of cardiovascular events anywhere between 2013 and 2020. Angiotensin-converting enzyme inhibitors, angiotensin II receptor blockers, beta-blockers, calcium channel blockers, and thiazides are the main drug classes of antihypertensives used in the prevention of CVEs [28, 29]. Lipid modifying agents is the main drug class in antihyperlipidemic monotherapy to prevent CVEs [28, 29]. Therefore, we defined antihypertensive monotherapy as receiving one of the following medications: thiazide-like diuretics with Anatomical Therapeutics Classification (ATC) code C03AA, aldosterone antagonists (C03DA), beta-blockers (C07A), calcium channel blockers (C08C, C08D, and C08E), angiotensin converting enzyme inhibitors (C09A), or angiotensin II receptor blockers (C09C). And we defined antihyperlipidemic monotherapy as receiving one of the following medications: HMG CoA reductase inhibitors (C10AA), fibrates (C10AB), bile acid sequestrants (C10AC), nicotine acid and derivatives (C10AD), or other lipid modifying agents (C10AX).

2.4 Inclusion and Exclusion Criteria

We selected patients in the claims database who were 19 years or older at the index date (defined as the first antihypertensive or antihyperlipidemic drug prescription), to select patients that were eligible for the deductible in the whole calendar year. Patients must have received at least three antihypertensive or antihyperlipidemic drug prescriptions within a year and have had at least 2 years of history in the claims database before they started treatment in the study cohort. The index date was the starting date of the follow-up defined as the first date of a prescription start. Patients that started both antihyperlipidemic and antihypertensive monotherapy at the same time are excluded.

Patients who were on secondary prevention, had at least one prescription of vitamin K antagonists with ATC-code B01AA, platelet aggregation inhibitors (B01AC), organic nitrates (C01DA), or other vasodilators used in cardiac disease (C01DX) or had a hospital claim related to cardiovascular events (Supplementary Table A1) in the 2 years before or 90 days after the index date were excluded.

Furthermore, patients that had at least one prescription of high ceiling diuretics (C03C), triptans (N02C), calcium, combinations with vitamin D and/or other drugs (A12AX), vitamin D and analogs (A11CC), or thyroid hormones (H03AA) in the 2 years before or 90 days after the index date were excluded to exclude patients with either chronic, stable heart failure, migraine, hyperparathyroidism, or thyroid problems. Patients that had at least one prescription of phentolamine (C04AB10) or tolazoline (C04AB02) in combination with at least one prescription of anticorticosteroids (H02CA), mifepristone (G03XB), or metyrapone (V04CD) were excluded, which was necessary to exclude patients with adrenal disease. The results of the inclusion and exclusion criteria are plotted in a flow chart.

Patients were followed until they left the health insurance company, had at least one prescription used to define a cardiovascular event including vitamin K antagonists (B01AA), platelet aggregation inhibitors (B01AC), organic nitrates (C01DA) or other vasodilators used in cardiac disease (C01DX) [30], had a hospital claim related to cardiovascular events (Appendix A), had died, or reached the end of the database, whichever came first.

2.5 Outcome

The coprimary outcomes were the adherence to antihypertensive medication and the adherence to antihyperlipidemic medication. Adherence was defined as a continuous number between 0 and 1 (both 0 and 1 included), calculated using a sliding window Continuous Multiple interval measure of medication Availability (CMA) with a window of 90 days [31]. Adherence was measured every 90 days from the index date until the end of follow-up for both antihyperlipidemic and antihypertensive medication.

2.6 Exposure

The exposure was defined as reaching the deductible limit. Reaching the deductible limit in a year was handled as a time- and patient-dependent variable and is set to “not reached” at the start of a new calendar year (1 January).

2.7 Covariates

We used the modified disjunctive cause criterion as described by VanderWeele [32] to select those covariates that are causes of either the exposure (reaching deductible limit), the outcome (adherence), or both and are available in the database. We assume that age, sex, social economic status, income, education level, selected comorbidities, and insurance choices, such as extra health insurances and insurance policy information (such as being the main insured, being collectively insured), are possible causes of the outcome and/or the exposure.

Information about age, sex, social economic status, income, education level, and other information defined on the insurance policy was collected at the index date. All insurance policy-related information was repeatedly collected at the start of a following year (1 January). Patients that had at least 180 defined daily doses (DDD) of blood glucose lowering drugs, with ATC codes starting with A10, within a year or had a hospital claim related to diabetes, were flagged as having diabetes from the moment of their first prescription or hospital claim date (Supplementary Table B1), whichever came first. In the same way we created a covariate indicator for asthma or chronic obstructive pulmonary disease (COPD). This indicator is 1 when a patient received at least 180 DDD of glucocorticoids (R03BA), adrenergics in combination with corticosteroids, or other drugs excluding anticholinergics (R03AK) or adrenergics in combination with anticholinergics including triple combinations with corticosteroids (R03AL) within a year or had a hospital claim related to asthma and/or COPD (Supplementary Table B1); the indicator was 0 otherwise. Patients were recorded as having rheumatoid arthritis when they had at least 180 DDD of medication listed in Supplementary Table B2 within a year or as hospital claim related to rheumatoid arthritis (Supplementary Table B1). Identification using medication and hospital claims is based on the pharmacy-based cost groups and diagnoses-based cost groups in the Dutch risk equalization system [33, 34].

Information about paid deductibles and copayments, such as the amount and for which medical service these are paid, were collected throughout the complete study period. Both the indicators for diabetes, asthma/COPD, and rheumatoid arthritis, as well as information about deductibles and copayments, were treated as time-dependent covariates. Furthermore, we added the covariate antihypertensive user, divided into the classes “adherent user,” “non-adherent user,” and “non-user” to the antihyperlipidemic model, where an “adherent user” is defined as having an adherence of more than 80% and a “non-user” is defined as using no antihypertensive medication (anymore). Similarly, we added the covariate antihyperlipidemic user to the antihypertensive model.

2.8 Statistical Analysis

First, a summary of all variables was made, including mean and standard deviation for continuous variables and a frequency count for categorical variables. To identify the association of deductibles with the outcome while correcting for other covariates, we used an ordered beta regression mixed model with a logit-link function and autoregressive covariance structure using the glmmTMB package [35] in R to model the adherence as a continuous number between 0 and 1. We investigated antihypertensive and antihyperlipidemic monotherapy separately. We used a mixed model to account for repeated measurements and within patient variation. The exponent of the β coefficients using this model were interpreted as the ratio in formula (1) and were referred to as the relative adherence ratio (RAR). This is similar to the definition of an odds ratio [36].

$$\begin{array}{c}{e}^{\beta }=\frac{{\mathbb{E}}\left[{\mathrm{Adherence}}_{\mathrm{new}}\right] / \left(1-{\mathbb{E}}\left[{\mathrm{Adherence}}_{\mathrm{new}}\right]\right)}{{\mathbb{E}}\left[{\mathrm{Adherence}}_{\mathrm{old}}\right] / \left(1-{\mathbb{E}}\left[{\mathrm{Adherence}}_{\mathrm{old}}\right]\right)}=RAR\end{array}$$
(1)

Adherence in the new setting could be for instance having reached the deductible limit, compared with the old setting were the deductible limit is not reached yet, keeping all the other factors constant [36]. Furthermore, we plotted the difference in expected values for binary deductible related variables in our dataset to show the direct impact on adherence, similar to an average treatment effect (ATE).

3 Results

After applying inclusion and exclusion criteria, we selected 30,565 patients who started primary preventive antihyperlipidemic monotherapy and 75,460 patients who started antihypertensive monotherapy anywhere between 2013 and 2020 (Fig. 1 and Table 1). Patients were followed for a median of 4.1 [mean 4.4, standard deviation (SD) 2.3] years. Whereas almost all variables were evenly distributed across the two different medication groups, the percentage of males and prevalence of diabetes was increased in the group that started primary preventive antihyperlipidemic monotherapy, compared with the group that started primary preventive antihypertensive therapy. Figure 2 shows the percentage of patients on antihypertensive or antihyperlipidemic therapy that reached their deductible limit after the index date. Since the index date for every patient is not at the same point in time of a calendar year, we see both an increasing as well as a seasonal effect in this plot. The distribution of other variables at index date can be found in Supplementary Table C1. In Fig. 3 we see a decreasing mean adherence over time for all patients that started either primary preventive antihyperlipidemic or antihypertensive monotherapy, for both patients that reached the deductible limit and those patients who did not. The overall adherence of patients that started primary preventive antihyperlipidemic monotherapy was lower than the adherence of patients who started antihypertensive monotherapy. Due to a small number of patients, the graph of the adherence of antihyperlipidemic monotherapy shows a steep decrease and increase toward the end of year 7.

Fig. 1
figure 1

Results of applying inclusion and exclusion criteria on the database such that only patients on primary preventive antihyperlipidemic or primary preventive antihypertensive therapy are selected

Full size image
Table 1 Characteristics at index date of patients on primary preventive antihyperlipidemic (N = 30,856) or antihypertensive monotherapy (N = 75,460)
Full size table
Fig. 2
figure 2

Percentage of patients on antihypertensive or antihyperlipidemic therapy that reached their deductible limit

Full size image
Fig. 3
figure 3

Left: mean adherence of patients on antihyperlipidemic monotherapy split according to patients who exceeded the deductible limit and those that did not. Right: mean adherence of patients on antihypertensive monotherapy split according to patients who exceeded the deductible limit and those that did not

Full size image

Using the multivariable time-dependent mixed model we found that, patients on antihypertensive and antihyperlipidemic monotherapy, those that reach the deductible limit are slightly (barely significant) more adherent than those that have not yet reached the deductible limit, with a relative adherence ratio (RAR) of 1.03 (95% CI: 1.00–1.05, p = 0.03) for antihyperlipidemic therapy and 1.02 (95% CI: 1.00–1.04, p = 0.05) for antihypertensive therapy. Those patients who pay the complete mandatory deductible in installments compared with those who do not pay their deductible in installments have an increased adherence to antihyperlipidemic and antihypertensive therapy (RAR 1.42, 95% CI: 1.28–1.57, respectively RAR 1.59, 95% CI: 1.46–1.73). Patients that choose an extra voluntary deductible (increasing the deductible) have a decreased adherence compared with those who did not choose an increase in their deductible (Fig. 4). Patients with a reinsured deductible (effectively no deductible) have an increased adherence compared with those without a reinsured deductible (Supplementary Figs. S1–S4).

Fig. 4
figure 4

Left: predicted mean adherence of patients that started antihyperlipidemic monotherapy with and without an extra voluntary deductible. Right: predicted mean adherence of patients that started antihypertensive monotherapy with and without an extra voluntary deductible

Full size image

Adherence to antihyperlipidemics is increased in adherent antihypertensives users compared with non-antihypertensive users with a RAR of 3.18 (95% CI: 2.96–3.41). The adherence of antihypertensives is increased in adherent statin users compared to non-statin users (RAR 5.87, 95% CI: 5.52–6.24). The association of non-adherent users of antihypertensives and statins between the adherence of antihyperlipidemic respectively antihypertensive therapy compared with non-users is inconclusive (Fig. 5).

Fig. 5
figure 5

Left: expected mean adherence of patients that started antihyperlipidemic monotherapy, being a non-antihypertensives user, a non-adherent antihypertensives user, or an adherent antihypertensives user. Right: expected mean adherence of patients that started antihypertensive monotherapy,were non-antihyperlipidemics users, were non-adherent antihyperlipidemics users, or were adherent antihyperlipidemics users

Full size image

Furthermore, for every year a patient is older at the index-date the adherence increases for both antihyperlipidemic as well as antihypertensive therapy (RAR 1.03, 95% CI: 1.02–1.03 and RAR 1.11, 95% CI: 1.11–1.12, respectively). Women have a lower adherence than men for antihypertensive therapy (RAR 0.70, 95% CI: 0.63–0.78), while there is no difference between men and women for antihyperlipidemic therapy (RAR 0.96, 95% CI: 0.85–1.10) (Supplementary Fig. S5). Being diagnosed with diabetes increases the adherence to both antihyperlipidemic and antihypertensive monotherapy (RAR 2.55, 95% CI: 2.31–2.81; respectively RAR 3.63, 95% CI: 3.16–4.16). Diagnosis with asthma/COPD has a statistically significant positive association with both the adherence to antihyperlipidemic therapy, as well as the adherence to antihypertensive therapy. Diagnosis of rheumatic arthritis has a negative association with adherence to antihyperlipidemic therapy, while the association with antihypertensive therapy is statistically non-significant (Supplementary Figs. S6–S8). All other RARs and corresponding p values can be found in Table 2 and Supplementary Table C2.

Table 2 Adjusted relative adherence ratios (RAR) for both antihyperlipidemic and antihypertensive monotherapy
Full size table

4 Discussion

We observed four different associations of deductibles in adherence rates. First, people that have reached the deductible limit in a year are more adherent, although a barely significant result, than those that have not (yet) reached the deductible limit. The result coincides with a study by Schneeweiss et al. [11], who concludes that out-of-pocket payments for statins is associated with non-adherence. Second, patients that have reinsured their deductible are more adherent than those that did not. Note that a reinsured deductible in fact means that the patient does not have to pay his deductible anymore. Hence the method of reinsuring deductibles could be used to increase adherence of cardiovascular medication for people with a low income. Third, patients that pay their deductible in monthly instalments in combination with their premium are more adherent than those that do not. Monthly installments are mostly chosen by people that are unhealthier and know they have to pay their full yearly deductible beforehand [24]. These two associations indicate that getting rid of the deductible could increase adherence. Finally, people that choose to voluntarily increase their deductible were less adherent (although barely significant for antihyperlipidemic monotherapy), irrespective of whether they reach the limit or not. Patients might not be willing to pay for their medication, because of their higher deductible. Alternatively, this might be explained from differences in health. Earlier studies have shown that people with a voluntary deductible are healthier than people that choose not to increase their deductible [37]. That is, this effect might be either a result from patients who believe they do not need medication, or patients that are not willing to pay for their medication, because of their higher deductible. Therefore, allowing to increase the deductible could be direct positive financial result for the health insurer, while it could be an indirect burden for both the patient and, in the long term, for the health insurer. However, the percentage of patients with a voluntary deductible, reinsured deductible, or paying their deductible in monthly installments was quite low (Table 1). Overall, the results seem to point in the direction that the need to pay for medication (or decrease the feeling of needing to pay for medication) decreases the adherence to this medication; however, statistical significance is marginal, which indicates a need for further research.

Comparing the association between deductibles and the adherence of cardiovascular medication with the association of both copayments (28%) [13] and coinsurance systems, we conclude that deductibles are less associated with adherence [10,11,12,13].

We found an increasing adherence with increasing age for both antihyperlipidemic as well as antihypertensive monotherapy, which is also observed in a meta-analysis in primary preventive statin therapy and a few studies in antihypertensive therapy [38,39,40]. Our finding that women were less adherent of cardiovascular medication is consistent with several previous studies [38,39,40,41].

An increased adherence in patients with diabetes compared with other patients, agrees with a meta-analysis of statin adherence by Hope et al. [41]. Although all included comorbidities seem to have a positive effect (with the exception of rheumatic arthritis) on the adherence of hypertensive monotherapy, this is in contrast with findings from some studies where comorbidities in antihypertensive therapy were found to decrease adherence [42,43,44].

We observed a decreasing adherence rate for both antihypertensive medication as well as lipid-lowering agents over time after initiation of therapy. The decreasing adherence of lipid lowering medication coincides with multiple studies on statin adherence [45,46,47,48,49]. Note that the rate of decrease is fairly similar to two studies conducted in the Netherlands despite using a (slightly) different population [46, 48]. The view on decreasing adherence over time of antihypertensive medication is also in agreement with several studies on adherence of antihypertensive medication, although not all studies are conducted in a population of primary preventive patients [49,50,51].

4.1 Strengths and Limitations

One of the strengths of this study is that it makes use of real-world data in combination with almost complete [except for over the counter (OTC) medication] medical records of patients. Furthermore, we used a mixed model to account for repeated measurement and within-patient variation. Also, we are able to distinguish between antihypertensive and antihyperlipidemic therapy and compare the results between the adherences of those two therapies. Importantly, the associations of deductibles with cardiovascular medication adherence if statistically significant were all in the same direction for both type of therapies.

This study is limited in that it does not measure the direct effect of being obliged to pay for medication, but uses a proxy of having reached the deductible limit. Also, observational studies similar to this one can only show associations and direct causal relationship should not be inferred. Reaching the deductible is impacted by adherence since using more medication leads to reaching the deductible earlier in the year. In addition, having reached a deductible also probably indicates higher use of healthcare generally. We adjusted for multiple comorbidities and other covariates, but these issues and others may represent residual confounding that is not fully controlled.

Another limitation is that for comorbidities, we had to rely on indirect measures, using medication and hospital records. However, we verified our cohort by comparing prevalence rates of comorbidities with those in other studies. The presence of asthma/COPD was similar to prevalence rates of asthma/COPD reported in another primary preventive antihyperlipidemic cohort [52, 53]. And similar prevalence rates for COPD/Asthma and rheumatic arthritis is found in the Dutch population [54, 55]. However, the prevalence of patients with diabetes at index date was lower than reported in other studies in primary preventive statin users [52, 53, 56]. Regarding primary preventive antihypertensive therapy, we also measured a lower prevalence of diabetes at index date than in most other studies [57,58,59]. The reasons for this are unknown. But we did find a bit higher overall prevalence than the prevalence of diabetes medication users in the Netherlands [54].

5 Conclusion

We measured a small negative effect of the need to pay deductibles in health insurance on the adherence of both preventive antihypertensive therapy and preventive antihyperlipidemic therapy independent of other risk factors for non-adherence. Other deductible related variables, such as a voluntarily increase of the deductible or a reinsured deductible were more strongly related to a decrease in the adherence of both antihyperlipidemic and antihypertensive monotherapy. Because of the complexity in causality of the effects, further study is needed on the potential health-economic consequences.