Preferences for innovations in healthcare delivery models in the Swiss elderly population: a latent class, choice modelling study

Abstract Background With the increasing number of people affected by multiple chronic conditions, it is essential for public-health professionals to promote strategies addressing patient needs for coordinated care. We aim to explore preference heterogeneity for better-coordinated care delivery models in Swiss older adults, and identify profiles of individuals more open to healthcare reforms. Methods A DCE (discrete choice experiment) survey was developed online and on paper for the Swiss adults aged 50+, following best practice. To elicit preferences, we estimated a latent class model allowing grouping individuals with similar preferences into distinct classes, and examined what background characteristics contributed to specific class membership. Results The optimal model identified three classes with different openness to reforms. Class 1 (49%) members were concerned with premium increases and were in favour of integrated care structures with care managed by interprofessional teams. Individuals in class 2 (19%) were younger, open to reforms, and expressed the needs for radical changes within the Swiss healthcare system. Class 3 respondents (32%) were strongly reluctant to changes. Conclusions Our study goes beyond average preferences and identifies three distinct population profiles, a majority open to reforms on specific aspects of care delivery, a smallest group in favour radical changes, and a third strongly against changes. Therefore, tailored approaches around healthcare reforms are needed, e.g. explaining the role of interprofessional teams in coordinating care, electronic health records and insurance premium variation.


I
ncreasing number of people with multiple chronic conditions poses considerable challenges in care organization, planning, and costs for public-health policies and healthcare systems worldwide.With the epidemiological transition from acute to chronic conditions and an increasing life expectancy, public-health professionals and policymakers must respond to a growing number of complex patients with frequent transitions between care settings.In Switzerland, chronic conditions account for more than 80% of total outpatient healthcare expenditures. 1oreover, with continuously increasing healthcare expenditures (>12% of GDP spent on healthcare), the country is facing increasing burden on households due to high out-of-pocket payments and community-rated premiums, [2][3][4] while healthcare resources are not always allocated efficiently. 1,2Therefore, a recurring topic of political discussions is the enhancement of equity in financing and the development of alternative models of healthcare delivery (e.g.6][7] With such models, care is provided by multiple healthcare professionals, requiring good communication and information exchange (e.g.access to patient history, care plan, etc.), aiming to deliver holistic and appropriate services reflecting patients' needs. 8,9In a recent expert consultation at the Federal level, care coordination was broadly discussed, and several coordination-improving measures were suggested: definition of patient pathways, enabling all providers to communicate better and develop a common understanding of their role; or promotion of networks with direct financial incentives for better coordination. 6However, implementing countrywide innovations may be challenging due to the complex, decentralized and fragmented organization and financing of the Swiss healthcare system. 2,3,10The complexity lies in the governance structure, whereby general healthcare regulations are imposed at the national level, but many more specific decisions are taken at the cantonal level (26 cantons).In addition, health insurance is mandatory in Switzerland but there is high private actor involvement in financing, as the plans are sold by many private companies (>50) in a strongly regulated market.Additional details on the Swiss healthcare system can be found elsewhere. 8n additional challenge to implement large-scale reforms is the Swiss political system that relies on direct democracy.The Swiss population can veto or demand reform through public referenda, 2 allowing populations to be directly involved in decision making.Therefore, an efficient communication strategy with clear explanations for the population is needed, as most voters may not understand proposals on complex subjects such as healthcare.In Switzerland, where development of innovative health policy implementations has been under heated debates for a long time, 3,11 it is critical to identify the population groups that would be willing to accept changes, and to tailor the communication strategies accordingly.
In such context, eliciting population preferences is a key element to increase acceptability of potential policy changes. 124][15] A DCE asks participants to repeatedly indicate their preferred option among hypothetical scenarios, and after completing a series of trade-offs the researchers can elicit preferences.Among the advantages of the DCEs are the ability to mimic decision-making processes in the real-world settings, visual attractiveness and interactivity for respondents, and relative ease of completion. 16There is an increasing trend of DCE use for the development of health policies by analysing the preferences of target groups for different aspects of the reform or intervention, as it allows to model the uptake of new initiatives or inform on the relative importance of aspects comprising a complex scenario. 14,16,17n the Swiss context, earlier DCE studies quantified trade-offs of the public and service providers between current healthcare insurance plan (status quo) and proposed innovations. 18,19][20] However, the importance of understanding how health treatments, policies or services could be tailored to various population groups has been widely recognized but still understudied, as their opinions and values are not homogenous. 21Implementing general recommendations following a 'one size fits all' approach may be inefficient and infeasible, 22 and the interest in exploring population heterogeneity has been growing rapidly among policy makers. 13,23pecifically, in the Swiss context where population is directly involved in the decision making via direct democracy, it is crucial to understand what population subgroups are more open to reforms, and what aspects of the reforms are of higher importance for them.
Therefore, the aim of this study was first to explore heterogeneity of population preferences for alternative better-coordinated models, and then to understand the profiles of Swiss adult population groups who were more open to healthcare reforms.

Selection of attributes
Details on the qualitative development of the survey, attributes and levels are available elsewhere. 22In brief, an extensive literature search was first performed yielding an initial set of 33 attributes, followed by multiple rounds of stakeholders' involvement, to obtain a manageable subset of attributes (N ¼ 8).These attributes were tested and refined in focus groups with the general population and patients.A final list of six attributes was validated in an online pilot study (N ¼ 301) and includes: access to EMR (electronic medical record), designated care coordinator, access to the specialists, formal compensation for informal care givers, exemption of chronic patients from paying deductibles and/or co-payments, and monthly premium change.

Survey design
The survey contained three main sections: general introduction and collection of background information, the experimental part (i.e. the DCE), and follow-up questions on health, healthcare use, choice of health insurance, and opinions on the Swiss healthcare system.The details of the study design may be found elsewhere. 8,22The experimental part of the survey consisted of repeated choices between two alternatives and a status quo option, and was presented in a 'dual response format'. 24Specifically, participants were first asked to choose between two hypothetical care models, and next to repeat their choice between the stated preferred option and a status quo ('my current model').
We used a D-efficient design and divided a subset of 42 possible choice sets into seven survey versions with six choice questions per version, which were used for subsequent choice modelling, plus one practice task and one consistency test task, which were used to evaluate the quality of the choice data.Details of experimental design are described elsewhere. 22An example of a choice set, and a survey version are presented in Supplementary Material.

Data collection
In total, 3472 individuals from the Swiss population aged 50þ and residing in French-speaking cantons were invited to participate by mail.We focused on population aged 50þ as this subgroup is characterized by the onset of multiple chronic conditions associated with increasing costs, and thus, they are more likely to benefit from new better-coordinated care models.Moreover, as such coordinated models do not exist in its defined form in Switzerland yet, it is important to understand preferences of the target population for the potential of such models, to inform their future design and communication strategy.For respondents under 70 years old, an invitation letter to participate in the online survey, programmed in Qualtrics (Provo, UT, USA), was followed by two reminders.The same procedure was used for potential participants aged 70 years and more, with an option of answering on paper upon request.A telephone hotline was set up to handle participants' questions and issues with the survey.Participants were rewarded via a lottery to win a CHF 300 cash prize (ca.USD 300).

Statistical analysis
First, we produced descriptive statistics of study participants: sociodemographic data, health status, healthcare utilization, insurance coverage and opinions on the Swiss health care system.For the DCE section, the outcome of each of the six choice sets was a three-level categorical variable (alternative 1, alternative 2, or current model [status quo]), which can be interpreted as the conditional probability of preferring one alternative over the others given the attribute levels in the choice set. 25o investigate preference heterogeneity, we estimated a latent class logit (lclogit, Stata 16.0) model allowing identification of individuals who can be grouped together. 26The optimal number of classes is exploratory, not predetermined, and is based on a measure of class differentiation for a range of possible class numbers (in our case 2-7), information criteria AIC (Akaike Information Criteria) and BIC (Bayesian Information Criteria).The model with the lowest AIC/BIC was considered superior to the other models (Supplementary table S1).To obtain a quantitative measure of how well the model did in differentiating several classes of preferences, we computed the average (over all respondents) of the highest posterior probability of class membership. 26ext, we described the classes by examining which background characteristics were associated with the probability of a certain class membership.First, each respondent was classified into the class with the highest posterior membership probability.Then, we analysed the associations between class membership and individual characteristics (e.g.socio-demographic, health-related, insurance-related, trust and opinions on Swiss healthcare, and health literacy) using multinomial logit model (mlogit, Stata 16.0).

Sample characteristics
Data were collected between March and April 2021.A total of 1385 individuals participated in the survey, 227 surveys were filled in paper format and 1158 online (response rate of 39% for individuals below 70, and 41% for individuals above 70 years old).However, 187 invitations were returned by the post office, e.g. because the recipient could not be found, or the letter was refused.Excluding the respondents who only partially or not at all completed the experimental part, we ended up with 975 fully completed surveys used for analysis, including 118 surveys in paper format.The median time for online completion was 23.9 min; overall 82% passed the consistency test, and 21% systematically chose the status quo option.Median age was 65, there were more males (59%) than females, and 36% reported having 0 diagnosed chronic conditions, while 32.3% reported having been diagnosed with 2þ diseases (table 1).
Preference heterogeneity for healthcare delivery models in Switzerland 261  Preferences for healthcare models expressed by different population subgroups

Estimates from latent classes
The estimation procedure led to a 3-classes solution, with the following distribution of respondents across classes: class 1 (49%), class 2 (19%) and class 3 (32%).The mean highest posterior probability of class membership was 0.92, meaning that the model performed well at distinguishing different underlying taste patterns for the observed choice behaviours.In line with our previous findings, 8 respondents attached more importance to monthly premium, care coordination, access to EMR and specialist access.However, despite specific similarities across classes (e.g.disutility of premium increase or restriction of access to specialist), there was substantial heterogeneity in preferences across the three classes (table 2).Class 1 members were rather cost-concerned (regarding premium increase), and at the same time were in favour of more integrated care structures where care was not managed by one GP only, but rather by a team of professionals.It was shown by their strong preferences of extended access to EMR to all healthcare professionals involved in care (but not all doctors), as well as care coordination performed by an interprofessional team.Gatekeeping, restrictions of access to EMR, and having no designated care coordinator were valued negatively in Class 1 (table 2).Additionally, the members of this class were the most critical towards maintaining cost-sharing for chronic patients, while the attribute of informal care compensation did not seem to influence their choices.
Individuals in class 2 seemed to be the most open to changes in healthcare delivery (table 2).These individuals were most in favour of informal care compensation, had lowest aversion of gatekeeping plans, and were the least responsive to premium changes (only the increase in premium was valued negatively).Other than that, they preferred to have a GP or a team as the care coordinator and extended access to EMR to all professionals or doctors involved in care.
Class 3 respondents had the highest probability of choosing the status quo; they showed reluctance for any attributes diverting from the current situation: e.g.strongest aversion of premium changes both negative and positive, any models with restricted access to the specialists, and informal care compensation.It needs to be noted, that the standard errors in this class were rather high reflecting a high level of randomness in the responses.Additionally, the members of Class 3 preferred the chronic patients being fully exempted from paying the deductibles together with co-payments, and negatively valued only partial exemptions (only co-payments, or only deductibles) (table 2).

Class membership description
Compared with Class 1 (majority of respondents), members of Class 2 were younger, higher educated, less wealthy, and were more likely to have low deductibles and gatekeeping plans (table 3).The members of Class 2 were 39% more likely to provide informal care for their relatives.On the other hand, they were 16% more likely to report having one chronic condition and 20% less likely to be multimorbid than the members of Class 1. Noteworthy, the members of Class 2 contained 98% of all individuals who never chose the current situation (status quo option) in the DCE experiment, and related to that, were 12% more likely to express that Swiss healthcare system needed substantial reforms.Additionally, they were 26% more ready to pay higher taxes to improve the healthcare system, and 26% more likely to express that the system when the richest can afford better healthcare than poorest is unjust (table 3).
The most distinguishing characteristics of membership in Class 3 is that it contained 100% of all those who consistently chose the current situation.Compared with Class 1 (majority), the members of Class 3 were likely the most senior, less wealthy and lower educated (table 3).Moreover, the members of Class 3 were 27% less likely to express that Swiss healthcare system needs large reforms and 24% less likely to be in favour of a public fund introduction for managing obligatory health insurances.

Discussion
In the present study, we demonstrated that analysing preferences from the perspective of an average individual is not optimal, and various subgroups of the population must be accounted for to develop and communicate healthcare reforms and innovations.We distinguished three distinct classes within our study population: the majority (49%) was open to some reforms mostly for integrated healthcare structures; a smaller class of adults most open to reforms (19%), and a class of conservative individuals, reluctant to changes (32%).
We found that the largest group attached high importance to a stronger role of interprofessional teams in coordinating care, access to the EMR and health insurance premiums.8][29] Specifically, in case of healthcare delivery policymaking, it is important to develop communication strategy and information campaigns reflecting the needs and interests of this group and emphasize benefits from better information exchange among the providers with possible access to electronic health records, the leading roles of coordinating healthcare professionals and measures to prevent premium increases.The findings that the largest population class prefers to switch away from the care centred around one general practitioner towards integrated care seem promising, as it coincides with the general agenda of Swiss policy. 6,30There is the significant scope to broaden the uptake of non-physician professionals in primary care in Switzerland, whereby the inclusion of new professional roles can lead to higher care quality and satisfaction, without affecting costs. 31econdly, we demonstrated that there is a group of most open to reforms individuals, who are younger in the adult population, healthier, express an opinion that the system where the existing healthcare system is unjust, and they are ready to pay more taxes for more-equity healthcare.Similar findings were observed in the USA, where respondents who felt that income-based healthcare inequalities were unfair, were more likely to support major health system reform, although much larger health and income disparities were found in the USA than in Switzerland. 32In an earlier study from Switzerland, it was found that the majority of the respondents were in favour of broad public-health reforms, 33 whereby those interested in politics expressed more support towards healthcare access-improving reforms.This population subgroup is more likely to actively participate in the political discussion in healthcare, and consequently, it would be beneficial to involve them into discussions about the upcoming reforms to reach greater patient involvement.
Finally, the last group of respondents were most senior, less wealthy, and less educated yet more conservative individuals, usually choosing the status quo and showed reluctance for any attributes Preference heterogeneity for healthcare delivery models in Switzerland 263 diverting from the current situation, especially limiting the freedom of provider choices.Individuals in this group expressed an opinion that Switzerland needs little or no healthcare reforms and were less likely to support any alternative means of financing healthcare (e.g.single fund or tax-based).Individuals of older age in Switzerland generally showed higher satisfaction with current situation in healthcare quality and healthcare policy, compared with younger adults 16,18,34 and to the other countries. 35Moreover, earlier studies from Switzerland showed that due to economic reasons, 18,20,36 the more senior and less healthy individuals were less likely to give up their current healthcare models.However, it is not clear whether the reluctance to any changes is related to general satisfaction with system performance, or the lack of up-to-date information about the current healthcare issues the country is facing.This would be the population subgroup benefitting most from the nation-wide communication strategies, focusing on the reasons for changes as the existing situation is suboptimal, and potentially healthcare education campaigns.
Our study makes an important contribution to the literature exploring preferences among Swiss population groups towards potential healthcare system changes to strengthen policy decisions.Although earlier studies and the history of the Swiss votes in healthcare showed strong attachment of the Swiss population to status quo, 2,3,[18][19][20] we were able to demonstrate that not all individuals are reluctant to changes.Current study's novelty is in demonstrating the preferences, shares and characteristics of population subgroups most open to healthcare reforms, accepting only specific reforms and those who are unwilling to accept changes.In the Swiss context, these findings are crucial to develop the tailored reforms and establish the effective communication strategy focused on the needs of specific population groups.Currently, as the alternative models of healthcare delivery for chronic patients are under heated debates, the results of our study would be of high relevance for policymakers, public-health professionals, health insurance representatives, researchers developing innovative policy plans, and engaged professionals in the discussion of potential adoption of healthcare reforms.We reached a relatively large sample of Swiss residents aged 50þ and reduced the selection bias occurring in the online sampling studies by applying dual age-adapted data collection mode. 8oreover, we involved multiple stakeholders together with the general population and patients into the development of the DCE study, from the very first stages resulting in potentially higher credibility and acceptability of the study results. 34Finally, the engagement of multiple actors allowed defining care delivery characteristics, previously not mentioned in the literature (e.g.informal care compensation or exemption of chronic patients from payments).
However, our study is subject to limitations.First, our sample consisted of an adult population residing in only Swiss Frenchspeaking regions, not representative of the general Swiss population.There are likely cultural differences in perceptions of healthcare in various language regions in Switzerland, 36,37 therefore, investigating language-specific preferences towards new models of healthcare may be of interest in the future research.We focused on individuals aged 50þ as they were the most likely to benefit from innovative bettercoordinated healthcare models, and were characterized by particularly pronounced attachments to status quo. 18Second, specific DCE methodological limitations need to be mentioned, such as concerns of external validity due to hypothetical nature of choices and its consistency with real-world situations, simplifying heuristics of respondents, and insufficient respondents' involvement into the tasks limiting the quality of the data.We tried to minimize such occurrences by testing the survey during multiple steps during focus groups and an online pre-test.Third, earlier studies revealed the concerns about the growing complexity of the DCE models, which not necessarily adds to the precision and robustness of conclusions 21 but rather risk overfitting data. 38We used a relatively accessible latent class analysis (LCA) model and limited the number of classes to an interpretable amount, so we were able to distinguish them well enough to answer the research question.

Table 1
Characteristics of the study sample