Principal findings
In the quantitative phase, we assessed the relative importance of patient characteristics with regard to individuals having forgone recommended healthcare due to cost-sharing payments. The regression model indicated that having a higher income reduced the odds of having forgone recommended healthcare due to the deductible (ORs of higher income categories relative to the lowest income category (reference): 0.29–0.49). The model also revealed several significant relationships across various determinants. Being older, the presence of one or more chronic conditions, having a higher level of mastery and more financial leeway were all shown to be protective factors (i.e., decreased the odds of having forgone recommended healthcare due to the deductible), while having a moderate or good self-reported health showed to be a risk factor (i.e., increased the odds). Dominance analyses revealed that financial leeway was the most important patient characteristic: this determinant contributed the most (34.8%) to the model’s overall R2mf (i.e., 0.123), followed by income (25.6%), age (19.6%) and sense of mastery (8.9%). Relative to the main model, the results of additional models stratified by type of healthcare service and of the population weighted models (i.e., IPW models) revealed no meaningful differences.
In the qualitative phase, we conducted interviews to understand and enrich the quantitative findings. Four main themes were distinguished that affected the patient’s decision whether to use healthcare: (1) financial barriers, (2) structural barriers related to the complex design of cost-sharing programs, (3) individual considerations of the patient, and (4) perceived lack of control regarding treatment choices within a given treatment trajectory. Furthermore, “having forgone healthcare” seemed to have some negative connotation as the topic of the interview had to be reframed using more neutral terms.
Possible explanations and comparison with the literature
Our quantitative findings indicating the importance of financial leeway and income, correspond with previous studies that have linked factors such as the price of a given healthcare service, available household resources and income to the response in demand for healthcare [3, 36, 37]. Our quantitative findings also correspond with our qualitative findings as analyses distinguished financial factors as a relevant theme. In addition, and in line with literature [3, 8, 11, 38], we found a stronger response to cost-sharing among low-income interviewees relative to those with higher incomes.
Moreover, dominance analyses revealed that financial leeway was more important than income. On the one hand, this implies that an individual who is able to save some money for future health expenses despite having a low income, is less likely to forgo healthcare due to these expenses, and vice versa. On the other hand, this finding reflects the impact that unexpected expenses (e.g., due to multiple cost-sharing payments) or a sudden drop in income (e.g., being self-employed with no clients) may have on an individual’s financial situation and, in turn, they forgo healthcare due to the costs involved. These findings correspond with our qualitative findings. As regular users of healthcare, interviewees often had to pay the full deductible. To minimize the impact of paying such deductibles on their financial leeway, most interviewees had arranged to pay by monthly installments. Having to pay the deductible in itself therefore played a minor role.
In line with literature [37, 39], our qualitative analyses distinguished the complexity of cost-sharing programs as a relevant theme, and also indicated that its relevance differed across educational level. Interviewees with a low to moderate educational levels had more difficulty in determining in advance whether, and if so, how much they had to pay for a given healthcare service. Unsure or unable to determine whether they could afford these costs, interviewees decided not to use the given healthcare service or stopped any future use. In contrast, the interviewee with a higher educational level was able to navigate effectively within the insurance plan.
Among the remaining determinants, being older, having one or more chronic conditions and having a higher level of mastery were protective factors, while having a better self-reported health level was a risk factor. It is reasonable to assume that, given their age and previous experience with the use of healthcare, older individuals and those with chronic condition are more likely to be aware of the potential adverse effects -that having forgone recommended healthcare may result in- compared to those who are younger or who have no chronic condition. Hence, they may be keener on maintaining their current level of health and thus be more incentivized to use the healthcare as recommended.
Regarding sense of mastery, our findings are in line with the literature: previous research has linked higher levels of mastery to better health levels, and suggests that those with such high levels are more capable (1) of effectively managing their health-related problems and (2) of using coping strategies to deal with these problems [40]. This mechanism also supports our qualitative findings as some interviewees had forgone healthcare because they had difficulties accepting their chronic conditions and the resultant problems.
Regarding self-reported health, our qualitative findings may provide an explanation. If the perceived medical benefits were too small considering their health level, interviewees would not use the given healthcare service despite their physician’s judgement. More specifically, having a better state of health may reduce the perceived medical benefits of the given healthcare service that, in turn, leads individuals to forgo healthcare. This post-referral consideration may also explain why many eligible individuals have declined to participate in our interviews as they would not classify themselves as individuals who forgo healthcare. Previous research has indicated that the term “having forgone healthcare” is often perceived as stigmatizing as it suggests that the individual acted irresponsibly and thus should be blamed for not using healthcare [41]. Hence, the seemingly rational consideration that individuals give to the matter after being referred, contradicts the common opinion that those who forgo healthcare are irresponsible.
Furthermore, as treatment trajectories are generally comprised of multiple health services, interviewees often perceived the use of health services as part of this trajectory as compulsory. Consequently, they may reconsider to follow up on a referral if they believed this could lead to a full treatment trajectory. Our findings are in line with literature. Lippiett et al. have shown in their review that treatment trajectories for lung cancer have been described as demanding in terms impact on everyday life (e.g., frequent hospital visits) [42]. According to Sav et al., when patients perceive the burden of treatment as high, non-adherence to treatment is the most likely consequence to occur [43].
Implications
Our findings have several implications. First, the observed importance of financial leeway indicates that solely adapting cost-sharing programs to income levels to prevent certain individuals from seeking recommended healthcare due to the costs involved (e.g., lower payments for low-income groups) will only get one so far. Individuals who are faced with multiple expenses due to frequent use of healthcare find that they are left with little financial leeway. To prevent such accumulation of expenses, policy makers need to adopt a broader perspective in which they consider all healthcare expenses that an individual may have at a given time and design their cost-sharing programs accordingly. Moreover, as cost-sharing payments reduce the demand for both recommended and non-recommended healthcare [8, 11], policy makers should follow the design principles of value-based health insurance that directly link these payments to the ‘value’ of the given healthcare service [44, 45]. More specifically, healthcare services that yield high value (i.e., substantial medical benefits for a patient’s health relative to their costs) should be subject to lower or no cost-sharing payments, while those with little value should be levied with higher payments. Policy makers should also consider the administrative costsFootnote 3 involved [46]. An all-payer claims processing data infrastructure, as implemented in the Netherlands, may help to limit administrative costs. In the Netherlands, all invoices between hospitals and payers are sent to and processed by a nation-wide system (i.e., VeCoZo). Processing individually adapted cost-sharing payments through the same nation-wide system should help to reduce the administrative burden.
Second, the relevance of the complexity of cost-sharing programs warrants additional efforts aimed at improving the transparency of these programs. For example, relative to a front-end deductible, flat-fee copayments paid at point of care offer individuals clear and immediate information on the required payments in advance; in a hypothetical decision context, Salampessy et al. [22] have demonstrated that such payments stimulate adherence to recommended healthcare.
Third, policy makers and physicians should be aware that various personal considerations and the perceived compulsory use of healthcare play a role in whether an individual uses healthcare. It underlines the importance of shared-decision making; a process that Elwyn et al. [47] have defined as “an approach where clinicians and patients make decisions together using the best available evidence” (p971). Policy measures that improve patient-centered care in clinical practice may help physicians to address these issues during consultations.
Strengths and limitations
A strength of our study is the use of an explanatory sequential study design. The mix of quantitative and qualitative methods enhances the quality of our inferences and leads to a deeper understanding of our findings [48]. In addition, we followed principles of good practice in qualitative research [34]. For example, we sought feedback on the summary of the interview (member check) to improve the credibility of our findings. Also, we collected and analyzed data iteratively, and discussed the findings with multiple researchers; all of which improved the dependability and confirmability of our findings.
Certain limitations to our study should however be noted. With regard to the quantitative phase, our sample was not representative of the whole Dutch population. Although IPW models based on weighted representative sample in terms of age, gender and educational level produced similar results, we did not have population data for other relevant characteristics such as health and sense of mastery. However, as our sample consisted of regular users of healthcare who have faced cost-sharing payments, their observed responses may resemble their decision behavior in real-life settings more closely, which improves the internal validity of our findings.
With respect to the qualitative phase, we did not achieve data saturation due to the small number of interviews. Also, due to this small number we may have missed other relevant perspectives such as those of young people. Both aspects reduce the transferability and dependability of our findings [34]. While more interviews conducted among a wider sample is required to capture all relevant themes (i.e., a full-scale qualitative study) and achieve data saturation, we believe that our qualitative data is rich enough considering its explanatory purpose: most quantitative findings have been supported by one or more subthemes.
Furthermore, recent studies have demonstrated that factors related to the COVID-19 pandemic affect a patient’s decision to use health care: for example, Karacin et al. have shown that fear for COVID-19 has reduced the adherence to chemotherapy among patients with cancer [49]. However, as our data was collected prior to the COVID-19 pandemic, we could not consider the effects of this pandemic. It remains unclear to which extent a factor such as fear for COVID-19 would have affected our quantitative findings. Regarding our qualitative findings, we expect that fear for COVID-19 will be distinguished as an additional (sub)theme.