Do Expectations of Risk Prevention Play a Role in the Adoption of Smart Home Technology? Findings from a Swiss Survey

Abstract

Smart homes offer promising opportunities for risk prevention in private households, especially concerning safety and health. For instance, they can reduce safety risks by detecting water leakages quickly and support health by monitoring air quality. Current research on smart home technology predominantly focuses on usability, performance expectations, and cyber risks, overlooking the potential importance of risk prevention benefits to prospective users. We address this gap by utilizing data from a recent survey to construct a structural equation model. Our overarching hypothesis is that prevention benefits and comfort considerations positively influence adoption. The results confirm the relevance of comfort, as suggested by previous research. In addition, the results reveal significant prevention benefits in safety and health, which are positively related to technology expectations and the intention to adopt smart homes. Furthermore, newly included variables such as technology affinity and active aging lifestyle emerge as indicators of potential smart home users, extending the knowledge of user characteristics beyond traditional sociodemographic indicators. The findings contribute to filling a gap in the current risk and technology literature and are also relevant for smart home device manufacturers and risk and insurance practitioners looking to evolve their business models.

1. Introduction

Smart home (SH) technologies are becoming increasingly common due to their potential to improve various aspects of daily life at home [1]. They refer to internet-connected devices and services that transform living spaces into remotely managed, digitized, and automated environments [2]. These devices and services are commonly found in the areas of comfort, security, health, and convenience [3]. Examples include robot vacuum cleaners, smart locks and cameras, smoke alarms, leakage detectors, smart lighting, and thermostats. Zavialova [4] estimates that the number of smart homes—hence, homes equipped with one or more smart home devices—reached 360 million worldwide in 2023, which is expected to more than double in the next 5 years. From a risk-management perspective, SH is becoming a valuable tool for preventing and mitigating household risks [5]. Prevention involves proactive measures to minimize potential risks [6] and Internet of Things (IoT) technologies such as SH use data to gain insights for risk management. Recent research on SH prevention highlights several promising applications, mainly to improve safety and health. For example, SH helps prevent water leaks by monitoring, alerting, and controlling specific water pipes in and around the home [7]. It also helps reduce injuries at home by preventing falls, ensuring child safety, and detecting unusual movement patterns [8,9].

Although SH devices are widely used in households and can potentially reduce risks, their main benefits—from a customer’s perspective—are convenience, efficiency, and comfort for daily living [10]. Research has focused mainly on investigating performance expectations and usability [1]. Knowledge of what drives user interest in SH for risk prevention is lacking. Practitioners in the insurance industry describe difficulties in attracting customer interest despite the industry’s expertise in using technology to assess and mitigate risks [11]. Our study builds on prior research on smart home benefits and adoption [3,12,13], the use of IoT for risk prevention [11], and a data set by Iten et al. [14] to investigate the associations of prevention-related variables with SH adoption. Our research contributes to the technology-oriented field of SH research by shedding light on the interplay between technological advances, human perceptions and preferences, and data-driven risk mitigation. Answering this question is also essential for device manufacturers and the insurance industry, which have had limited success in developing prevention-oriented services [15]. There are also specific motives to examine the risk-prevention proposition for SH adoption in Swiss households. With an estimated adoption rate of one in four, compared to higher rates in Scandinavia and the UK, Swiss priorities lean particularly towards safety and health [4]. This focus differs from leading countries, where comfort and environmental aspects are equally important. Additionally, Switzerland’s relatively low homeownership ratio, in combination with high living standards, further highlights the need to understand local adoption motivations in greater detail [16].

Against this background, we investigate how and to what extent perceived preventive benefits in safety and health and broader considerations of comfort are related to the intention to adopt SH. In addition, our analysis includes various personality traits to characterize potential users. Using data from a comprehensive survey on the dynamics of SH adoption in Switzerland, we build a structural equation model (SEM) to capture the hypothesized effects on the intention to adopt SH, our principal variable. Based on a stepwise selection procedure that has tested 34 regression variables, we extract a set of significant variables related to the personal characteristics of a prospective SH user. To compute the model, we use the method of partial least squares and aim to explain the central relationships within it.

The results indicate that potential users perceive significant preventive benefits in terms of both safety and health. These benefits are positively related to the performance expectations of the technology and the increasing intentions to adopt SH. As suggested by previous research [1,17], we confirm that comfort considerations are critical to adoption. Furthermore, specific variables emerge as indicators of SH adopters, including affinity for technology and knowledge, sociodemographics, and active aging lifestyle. Based on technology acceptance research, we gain new insight into the value of prevention for future SH adopters and fill a gap in the rather technology-oriented research on SH. The findings also have profound implications for the risk and insurance practice’s approach to the SH technology landscape. We conclude that offerings need to balance both dimensions, convenience and prevention benefits, and would benefit from focusing on user engagement beyond technical capabilities. Additionally, the results indicate that a market strategy targeted at specific segments, such as homeowners or active agers, holds the most promise.

This article is organized as follows. In Section 2, we review the literature on SH prevention and adoption, which informs the development of our research hypotheses in Section 3. We construct the model in Section 4 and present the available data. The model results are shown in Section 5. In Section 6, we discuss our findings and conclude in Section 7.

2. Literature Review

In this section, we provide an overview of the relevant research literature and use it to formulate hypotheses for the subsequent analysis. We begin by analyzing the prevention use cases associated with SH technology. We focus on use cases related to the physical characteristics of a home (property), as prevention benefits associated with wearable technology are beyond the scope of our research. Next, we summarize the drivers of SH adoption and identify the perceived benefits and user characteristics that drive interest in the technology.

2.1. Smart Home Prevention Use Cases

Prevention is one of many risk-management tools. Prevention seeks to influence the frequency or severity of risk by altering its occurrence or consequences [6]. IoT technologies, such as SH, promise to improve risk management practices by using data to provide valuable information for risk reduction [18]. A key discussion revolves around whether these risk-reduction activities serve as complements or substitutes to insurance and, consequently, their impact on insurance demand. Recent evidence increasingly suggests that they act as complements, thereby reinforcing the sustainability and effectiveness of these preventive measures [19]. In the insurance industry, the preventive potential of IoT has been demonstrated primarily through telematics and wearables. Telematics, which focuses on mobility use cases, allows drivers to monitor their driving behavior and receive alerts and recommendations [20]. Wearables focus on health applications by continuously monitoring individual health parameters, promoting a proactive approach to well-being [21]. In both cases, suggested behavioral changes are incentivized to promote safer practices, reduce claim costs, and align insurers’ and policyholders’ interests [22]. Recent studies have demonstrated specific impacts on the choice of lower levels of coverage, indicating policyholders’ intention to engage in preventive measures to reduce the likelihood of a loss occurring [23]. Prevention activities also become more important for the insurability of certain risks, as climate change effects and the digital transformation of the insurance sector have highlighted their limits. Prevention is expected to play a vital role in managing these risks, ensuring that both insurers and policyholders are better prepared for potential threats [24].

SH also promises opportunities for risk prevention. Flückiger and Carbone [11] note that the capabilities of real-time safety enhancement by technology have led insurers to speculate about the potential to reduce exposure to household risk and vulnerability to risks typically covered by insurance policies. Several promising use cases have emerged internationally, particularly among insurers in the United States. An example is Hippo’s home insurance offering, which includes smart devices to mitigate water damage, fire, and unauthorized access. An example of a European insurance offering is Luko in France. The company started in 2018 with a strong focus on SH prevention but gradually reduced its IoT activities. Finally, in the summer of 2023, the company sold its entire policy portfolio. Enzo, a German InsurTech company, has recently adopted the prevention premise of SH to reduce water leaks in homes. A comprehensive literature review by Iten et al. [5] details the household risks that SH technology can effectively address. The authors identified several emerging threats, including privacy and cybersecurity risks, performance and dependency risks, and everyday household risks related to theft, fire, water, and health.

Table 1. Review of selected risks and SH use cases addressed by SH products in the safety, health, energy management, and cybersecurity domains.

Risks and SH Use Cases	SH Products	References
Safety domain
Unauthorized access
– Simulate presence – Door entry management – Intrusion detection – Activation of panic scenarios – Windows monitoring	Smart door lock, video doorbell, motion sensor lock, alarm system, smart light, window controller, camera	Sovacool and Furszyfer Del Rio [3], Acoca et al. [25], AXA [26], Blythe and Johnson [27], Feuerstein and Karmann [28], Kivimäki et al. [29], Sevillano [30], Tural et al. [31]
Fire
– Smoke and CO2 detection – Shutdown and sprinkler systems – Escape plan assistance	Smoke alarm, water sprinkler system, interactive fire escape plan, intelligent kitchen appliances	Sovacool and Furszyfer Del Rio [3], Acoca et al. [25], Sevillano [30], Tural et al. [31], Gielen et al. [32], Hsu et al. [33], Karemaker et al. [34], Saeed et al. [35], Salhi et al. [36]
Natural hazard
– Weather monitoring – Earthquake detection – Wildfire warnings	Weather station, seismic sensor, wildfire alert system, window controller	Azam et al. [7], Feuerstein and Karmann [28], Sevillano [30]
Water damage
– Leaks and flood detection – Humidity control – Freezing and bursts prevention – Emergency shut off	Leak detection sensor, flood sensor, humidity control system, smart water valve	Azam et al. [7], Feuerstein and Karmann [28], Sevillano [30], Davis [37]
Health domain
Cognitive impairment
– Dementia monitoring – Voice-control assistance – Cognitive support	Activity sensor, voice-activated assistant, digital assistant	Brims and Oliver [38], Carnemolla and Bridge [39], Liu et al. [40], Oyeleke et al. [41], Saragih et al. [42], Wrede et al. [43]
Frailty
– Family member care – Promotion of exercise routines – Nutrition patterns monitoring	Activity sensor, smart camera, exercise machine, intelligent kitchen appliances	Sovacool and Furszyfer Del Rio [3], Hsu et al. [33], Carnemolla and Bridge [39], Liu et al. [40], Crane et al. [44], Gómez-Portes et al. [45], Kracht and Staiano [46], Murri et al. [47], Romero et al. [48], VandeWeerd et al. [49], Welch et al. [50]
Injuries
– Falls and injuries prevention – Ensuring child safety – Unusual movements detection	Medical alert system, childproofing sensor, motion sensors, camera	Roberts et al. [8], Torres-Guzman et al. [9], Abbassinia et al. [51], Ambrens et al. [52], Ma et al. [53], McKenzie et al. [54], Miranda-Duro et al. [55], Nose et al. [56], Pech et al. [57], Wright et al. [58]
Energy management domain
Waste of resources
– Energy usage optimization – Automation of appliances – Water resources management – Protection against power surges – Solar energy utilization – Electric vehicles charging	Smart thermostat, smart plug, smart water valve, smart sprinkler controller, solar panel, electric car charger, intelligent heating/cooling system	Sovacool and Furszyfer Del Rio [3], Amin et al. [59], Fernández-Caramés [60], Jones-Garcia et al. [61], Psomas et al. [62]
Cybersecurity domain
Cybersecurity and privacy
– Technical treatment – Raising awareness – Knowledge dissemination – User empowerment	Not available	Al-Begain et al. [63], Ali and Hong [64], Buil-Gil et al. [65], Hammi et al. [66], Jacobsson et al. [67], Klobas et al. [68], Pecorella et al. [69]

summarizes the prevention use cases identified in the literature. We associate the risks and cases of SH use with existing SH products. The results are organized according to the specific risks they address in safety, health, energy management, and cybersecurity.

In the safety domain, SH prevention use cases focus primarily on common household risks [30]. For example, SH contributes to preventing unauthorized access risks by simulating presence during absence, efficiently managing door access, and detecting intrusions. These use cases are realized through integrating various SH products, including door locks, video doorbells, motion sensors, and lighting systems [3]. Fire prevention is another prominent area, which involves the detection of smoke and carbon monoxide, automatic shutdown mechanisms, and interactive assistance with escape routes. Saeed et al. [35] find that SH sensors can effectively detect fire and solve wide-known limitations of false alarms. Safety-related use cases also extend to the prevention of water leakage and the mitigation of natural hazards. These include monitoring, notification, and control capabilities for specific parameters affecting the home and its surroundings [7].

Health-related SH prevention use cases are difficult to distinguish from other IoT applications in daily life [70] and often involve wearables and home applications. Most of the research comes from studies of specific diseases or demographics [39]. These studies usually focus on older adults or people in care settings. An often-explored health-focused application is injury prevention, including use cases to prevent falls, ensure child safety, and detect unusual movement patterns. McKenzie et al. [54] find that using SH-based sensors and gamified learning experience on child safety at home can lead to improved safety knowledge and safety actions for participants, although loss to follow-up can be a limitation. Kivimäki et al. [29] find that the ultimate wish of older people’s safety premise with SH was to be able to live in their own home for as long as possible. Motion sensors, cameras, and medical alert systems are critical in these efforts [39]. In addition, health-related use cases extend to preventing frailty [49] and helping with cognitive impairment [43].

Energy management is a growing area of research, with SH recognized as a critical lever in the contribution of households to a solution to climate-related challenges [25]. Numerous use cases in this area aim to reduce inefficient consumption habits. Smart thermostats, plugs, water valves, and integrated heating and cooling systems enable monitoring and optimizing energy and water use, automating the routines of household appliances, and the localized production and consumption of solar energy [59]. Among others, Jones-Garcia et al. [61] found that smart applications in the household can improve resource management, making an example of a `smart bin’ introduced as an effective intervention for making routine household food-wasting practices visible.

Cybersecurity and privacy concerns represent significant emerging risks related to adopting SH technology [68]. Preventive measures are mainly derived from research in information security, focusing on failure and intrusion detection systems. Jacobsson et al. [67] find that most severe emerging cyber risks are derived from human factors and are therefore inherently complex to handle, suggesting that mitigation efforts should be part of the design phase of SH. In addition to these technical studies, other work focuses on the knowledge that SH users need to proactively reduce their exposure to cyber risks [67].

2.2. Determinants of Smart Home Adoption

The literature on technology adoption is critical to understanding the users’ perspectives on SH. This body of research provides information on the factors that shape an individual’s decision to adopt and use SH technology [71]. Technology acceptance models have evolved based on the seminal work of Davis [72]. For example, an overview of the technology acceptance model and the unified theory of acceptance and use of technology model applied in the SH context can be found in Iten et al. [14] and Marikyan et al. [2]. Empirical evidence focuses on three issues: SH service areas and their benefits to users; the factors related to the adoption of SH, with particular emphasis on the concept of general usefulness or performance expectation of the technology; and the study of specific personality traits that characterize potential SH users.

2.2.1. Service Areas and Benefits

The main SH service areas include comfort, energy management, health, and safety [1]. In the area of comfort, SH services aim to improve the lives of residents by simplifying daily tasks and giving them greater control over household appliances, thus providing additional comfort [73]. SH energy services focus on reducing energy consumption through continuous monitoring and automation, significantly contributing to sustainability efforts by optimizing energy use [74]. Health services address individual health and environmental information needs, often focusing on older adults or individuals with disabilities, to promote healthier and more independent living [31]. Safety services enable residents to strengthen home safety, prevent accidents, and minimize financial losses [27]. Moreover, these essential service areas provide the expected benefits of SH and have a critical impact on the take-up of the technology by the typical household [2].

2.2.2. Key Drivers for Adoption

Among the main drivers of the adoption of SH, performance expectation or perceived usefulness of SH plays a pivotal role in shaping individuals’ intentions to adopt [31]. As individuals assess the benefits associated with SH, their assessment of the usefulness of the technology becomes a critical factor in their decision-making process [75]. In studies of SH adoption, researchers often contextualize SH’s utility in increasing perceived productivity, efficiency, and overall effectiveness in performing daily household tasks [17]. As such, the focus is typically on convenience, particularly the automation of everyday household tasks. According to Li et al. [1], comfort considerations are generally emphasized as the primary driver of SH adoption. Few studies have examined preferences for benefits in other service areas; for example, Chang and Nam [12]. These authors compared the relative importance of all four service areas in driving adoption intentions. The results highlighted the importance of comfort as the main driver, followed by safety considerations. More recent work has examined additional factors such as enabling conditions and social influences [76,77], fun and enjoyment [17,78], the perceived value of investing in technology [31], and barriers and risks [79].

2.2.3. Prospective User’s Characteristics

Knowing the personal characteristics of a potential SH user can be helpful because it can indicate which individuals are more likely to recognize the benefits of SH [80]. However, the results in this area are mixed and sometimes contradictory. Younger adults tend to show higher adoption intentions than older adults [81]. However, studies by Shin et al. [82] and Klobas et al. [68] have observed higher adoption rates among older adults, particularly in SH health settings, where they are more willing to share personal data. The influence of gender on SH dynamics remains inconclusive. Sovacool et al. [83] suggest that the effect of gender is related to the underlying promise of SH, observing significant differences between a promise focused on entertainment and reduced housework. Higher income and education levels positively correlate with SH interest [68], although Chang and Nam [12] suggest that this effect may be related to the cost associated with technology. The essential characteristics of SH users are experience and affinity for technology, previous experience with SH [84], the awareness and knowledge of SH technologies [85], the ownership of other technologies [86], and the ownership and expertise of a smartphone [31]. All are associated with higher levels of adoption of SH. Furthermore, marital status [87], homeownership [87], and household size [31] were found to be related to SH adoption.

3. Hypotheses Development

To guide our subsequent analysis, we develop hypotheses that synthesize information from SH prevention use cases and factors influencing technology adoption. These hypotheses consider performance expectations, personal characteristics, and comfort, safety, and health benefits. Our approach aims to provide a comprehensive overview of the factors that influence the adoption of SH technology in the context of its promise for risk prevention in Switzerland.

Prevention use cases studied in industry and academia highlight the importance of addressing health and safety considerations. These use cases focus on safety-related applications, such as preventing fires and water damage, and health-related applications, such as preventing injuries and helping people with cognitive impairments. Given the particularly central role of safety and health research in SH prevention, we propose the following hypotheses.

(H1)

Safety and health prevention benefits and performance expectations correlate positively.

The literature pinpoints the relevance of safety benefits for performance expectancy, as they are known as the most popular SH devices and benefits can be most easily associated with them. A similar perspective trend can be seen for health benefits when looking from the perspective of older adults at the perceived usefulness of SH [1,13].

(H2)

Safety and health prevention benefits and SH adoption intention correlate positively.

Research also emphasizes the critical role of safety and health benefits in SH technology adoption intentions, particularly in supporting safe and independent aging [31].

Research on SH adoption has mainly focused on the comfort aspect of technology and its influence on shaping performance expectations and SH adoption intentions. By validating the role of comfort in the context of the prevention aspects explored above, we further strengthen its validity and importance. Therefore, we propose to test the following hypotheses.

(H3)

Comfort benefits and performance expectations correlate positively.

Comfort benefits focus on enhancing SH technology’s convenience and ease of use. Studies have shown that comfort is typically a driver of performance expectations, as these devices significantly improve users’ perception of the technology’s usefulness [75].

(H4)

Comfort benefits and SH adoption intention correlate positively.

Comfort considerations are generally emphasized as the primary benefit of SH adoption. Most studies on SH emphasize that users are more likely to adopt SH technology when comfort benefits around efficiency and control are perceived [1,12].

Performance expectation is one of the main drivers of adopting SH technology identified in the literature. This facet highlights how people evaluate the practical usefulness and benefits they expect from incorporating SH into their homes. As a critical component of the adoption decision process, we propose the following hypothesis.

(H5)

Performance expectation and SH adoption intention correlate positively.

The literature indicates that users’ expectations of the technology’s performance, including its reliability, efficiency, and ease of use, significantly influence their adoption decisions. This relationship has been validated several times, among others, in the standard works of Davis [72] as well as Venkatesh et al. [88].

Furthermore, the literature on SH adoption presents inconclusive results on personality traits contributing to increased interest in SH. Given these unknown relevant personal characteristics of a potential SH user, we hypothesize the following.

(H6)

Personal characteristics and SH adoption intention may correlate positively or negatively.

Personal characteristics, such as age, gender, income, education level, and technology affinity, influence the intention to adopt SH technology. Using the example of age, research has shown that these characteristics can positively or negatively affect adoption decisions. While younger adults often show higher adoption intentions [81], Shin et al. [82] and Klobas et al. [68] found higher adoption rates among older adults due to their increased willingness to share personal data in SH health settings.

4. Methods and Materials

This section provides a comprehensive overview of the model and data used to investigate the research question and test the hypotheses presented. Our approach focuses on three main elements: the SEM, whose visual representation also illustrates the hypothesized relationships; the methodology used to compute the model, which outlines the techniques used to estimate our model; and the data set selected to facilitate our analysis.

4.1. Data Analysis

We chose partial least squares SEM (PLS-SEM) as our primary methodology to investigate the hypothesized relationship. SEM studies are highly regarded in the social sciences for their practicality and effectiveness [89,90]. A significant advantage of PLS-SEM is its ability to handle different types of measures efficiently. This is particularly valuable when studying complex concepts that are not directly observable, often referred to as latent “constructs”, such as the benefits of SH technology, which are influenced by numerous specific factors and beliefs. Especially when those constructs are defined in a formative way, the indicators define the construct rather than reflect it. PLS-SEM offers more flexibility, as it does not impose the same stringent assumptions as comparable, covariance-based models [91]. In PLS-SEM, constructs can act as exogenous drivers or be considered endogenous by other variables. To fully reflect the nature of a construct, we use manifest variables called measures or “indicators”. In general, several indicators measure constructs that represent a particular aspect of a construct. However, for simplicity, we also use “single-items” to refer to constructs measured directly by a single observable variable, eliminating the need for multiple indicators in the modeling process.

Figure 1 shows the complete model thatwe are considering. In the graphic, the numbers from (1)–(7) denote the measurement models relating the latent constructs to their indicators, and (8) and (9) represent the structural models [92]. The complete set of variables, including the labels, their type, a description, the values taken, and the reference to the source in the original questionnaire (see [14] (Appendix A)), is reported in

Table 2. Overview of the variables used in the model.

Variable	Label	Type	Description	Values	Source
$COM$	Comfort	Exo	Broad comfort benefits	Five levels from strongly disagree to strongly agree
$BRE$	Burden relief	Ind	Reduce burden of household activities	”	G1.1
$HIO$	Home information	Ind	Provide information and control options	”	G1.2
$VEN$	Value enhancement	Ind	Maintain or increase property value	”	G1.3
$SAF$	Safety	Exo	Prevention benefits related to safety	Five levels from strongly disagree to strongly agree
$SOS$	Sense of safety	Ind	Make feel more safely	”	G2.1
$SBO$	Safety booster	Ind	Increase home safety	”	G2.2
$RPR$	Risk protection	Ind	Protect against risks at home	”	G2.3
$HEA$	Health	Exo	Prevention benefits related to health	Five levels from strongly disagree to strongly agree
$HMA$	Health maintenance	Ind	Take care of oneself and avoid doctor visit	”	G3.1
$HEM$	Health monitoring	Ind	Monitor easily health metrics	”	G3.2
$HEN$	Health encouragement	Ind	Motivate to behave healthier	”	G3.3
$APR$	Accident prevention	Ind	Help to prevent accidents and health risks	”	G3.4
$PEX$	Performance expectation	Endo	Usefulness unrelated to service areas	Five levels from strongly disagree to strongly agree
$ESI$	Everyday simplification	Ind	Simplify everyday household activities	”	H1
$HOM$	Home monitoring	Ind	Monitor effectively state or progress of home	”	H2
$AMO$	Activity motivation	Ind	Motivate to do activities that they do not like to do	”	H3
$TAF$	Technology affinity	Exo	Familiarity with technology usage in general	Five levels from low to high
$TEX$	Technology experimenter	Ind	Pleasure in trying new technologies	Five levels from strongly disagree to strongly agree	E1
$TPI$	Technology pioneer	Ind	First to try new technologies	”	E2
$TXT$	Technology expert	Ind	Skills in using the smartphone or tablet	Five levels from poor to excellent	E3
$KAP$	Knowledge and preference	Exo	Prior knowledge and preference for a service area	Five levels from low to high
$KLE$	Knowledge level	Ind	Level of experience in SH	Five levels from no to very good knowledge	A1
$CAP$	Convenience application	Ind	Preferences for sensors serving convenience	Five levels from dislike to like	B1
$HAP$	Health application	Ind	Preferences for mobile health device	”	B2
$AGE$	Age	Single	Age in years	45–90 years	A2
$GEN$	Gender	Single	Gender of the respondent	Female, male	A2
$FRA$	Frailty	Single	Frailty in certain everyday activities	No, yes	D2
$HOW$	Homeownership	Single	Main residence ownership	Renter, owner	C5
$CAL$	Cultural activity level	Single	Participation in cultural activities	Hardly ever, few times a year, 1-2x month, 1x week, >1x week	D6.1
$AIN$	Adoption intention	Endo *	Intention to adopt SH	Five levels from strongly disagree to strongly agree
$IUS$	Intended usage	Ind	Intention to use technology in the future	”	O1
$PUS$	Predicted usage	Ind	Prediction to use technology in the future	”	O2
$OUS$	Opportunistic usage	Ind	Intention to use technology when opportunity arises	”	O3

Note: The column “Type” uses the abbreviations “Exo” for exogenous variables, “Endo” for endogenous variables, “Ind” for indicator variables, and “Single” for single-item variables. The references in column “Source” refer to the original survey data defined in Iten et al. [93]. * Adoption intention $AIN$ is the principle variable of interest.

.

4.2. Data Collection

Our study uses data from a recent survey on SH adoption in Switzerland [93]. The survey aimed to assess established determinants of SH adoption and potentially relevant features and user characteristics from a risk-prevention perspective. A professional polling agency conducted the 2022 survey targeting individuals aged 45 years and older, and the survey was available in both German and French. The survey was designed according to the CHERRIES guidelines for online surveys and the reporting checklist can be found in ([14] (Appendix C)). The questionnaire featured 122 questions organized into categories such as personal characteristics, prevention benefits, SH adoption dimensions, and risks and costs. Further, it included an abstract SH scenario, including multiple examples of SH products available in Switzerland, to capture preferences for different prevention benefits. Participation was incentivized for successful completion. Additional sample quotas and control questions were defined based on the respondents’ age, gender, language, and SH knowledge level. Initially, 2553 responses were received, resulting in 1515 final observations after quotas and control. Iten et al. [14] (Section 3.1) described the survey’s development, structure, and operationalization. Based on the methodology laid out in the benchmark articles by Hair et al. [92], Becker et al. [94], and Sanmukhiya [95] in the field of PLS-SEM, we examined the data set in five key areas to assess its quality. Although this procedure did not reveal any problems with missing data, data distribution, and common-method bias, we identified 2 suspicious responses and 11 outliers that we removed from the sample (see

Table A1. Details on the examination of the survey data.

Examination	Description
Missing data handling	With fewer than 5% missing values for each variable, we applied the mean replacement technique where needed. Across all variables, there were a total of 48 missing values. The variable sense of safety ($SOS$) has the highest number of missing values (20 values, 1.3%).
Suspicious responses	The original survey included screening questions, quality checks, and a randomization process to reduce the number of suspicious responses. Upon examination of the distribution and variance of the responses, we excluded two participants over 90 years of age from the sample.
Outliers	Using the Mahalanobis distance, we reveal missing values in the indicators $IUS$, $PUS$, and $OUS$ related to the variable $AIN$. We exclude the 11 responses that show missing values in the 3 indicators.
Data distribution	The data distribution analysis for skewness and kurtosis reveals no critical values. We observe that the variable $GEN$ has a kurtosis of −2.003.
Common-method bias	Assessing the VIF values of the inner model and those obtained from the random variable approach, we find that all variables that appear in the final model show VIF values significantly below the 5.0 threshold and even below the more cautious 3.3 threshold.

in Appendix A). Therefore, the final sample used in this study consists of 1502 responses. For our analysis, we extract the relevant variables, comprising indicators on SH adoption intention (3 variables) and performance expectation (3 variables), 10 variables on the benefits of SH (excluding 4 variables related to fitness), as well as a comprehensive set of variables related to personal characteristics (34 variables).

In

Table 3. Distribution of the survey answers in the indicator variables for adoption intention ($N=1502$, values in %).

Intended Usage (IUS)		Predicted Usage (PUS)		Opportunistic Usage (OUS)
Disagree	35.6	Disagree	33.8	Disagree	31.7
Neutral	30.8	Neutral	26.3	Neutral	20.5
Agree	33.6	Agree	39.9	Agree	47.8

, we present the statistics on the responses related to the principal variable of interest, the adoption intention. The intention to adopt SHs was measured using three items: “I intend to use SH in the future”, “I predict I would use SH in the future”, and “If the opportunity presents itself in the near future, I will use SH”, with responses recorded on a five-level Likert scale. The three indicators of adoption intention show that slightly more people expressed an intention to adopt SH than those who did not. However, most of the respondents reported low levels of their knowledge of SH. Both observations are consistent with the patterns found in recent studies on SH adoption [78]. Iten et al. [14] provide descriptive statistics for all variables in the data set for the intention to adopt SH.

4.3. Methodology, Measurements, and Model Specification

4.3.1. Methodology

PLS-SEM uses a two-stage modeling approach that involves first estimating the measurement models, which describe the relationships between the observed indicators and the latent constructs (see Equations (1)–(7) below and Figure 1), and then estimating the structural models, which describe the relationships between the constructs and the single items considered (models A and B). PLS-SEM offers several advantages over the traditional covariance-based SEM (CB-SEM) approach, particularly in the context of our study. CB-SEM treats constructs as common factors that explain the covariance between their associated indicators, adhering to the “reflective” measurement philosophy, where indicators are manifestations of the underlying construct [96]. In contrast, “formative” measurements, where the construct is formed by its indicators, are better suited to the composite-based approach of PLS-SEM [91]. Further, PLS-SEM is better equipped for handling complex models, including formative and reflective constructs. It does not impose the stringent assumptions of CB-SEM, such as multivariate normality and large sample sizes, making it more flexible and suitable for exploratory research [97]. This flexibility aligns with the objectives of our study, allowing for the modeling of latent constructs as linear combinations of their indicators, which is essential for formative constructs [91].

In the first step, the indicator scores for a construct are combined to form a composite score using a linear weighting process [98]. In this process, Likert-scale responses are treated as ordinal variables and coded numerically using a scale ranging from 1 to 5. PLS-SEM can effectively handle ordinal variables, even when the numerical values do not represent equidistant intervals, as it focuses on the covariance structure rather than a specific distributional assumption [92].

In the second step, PLS-SEM estimates the path coefficients, i.e., the hypothesized relationships. Using ordinary least squares (OLS) regression and the standardized scores from the first step, the algorithm regresses the exogenous constructs on the endogenous constructs. The goal is to maximize the explained variance of the endogenous constructs, especially our principal variable, the intention to adopt SH. We measure the explained variance by the models’ R-squared values. PLS-SEM is the preferred method for developing theory and explaining variance [94].

In the following, we describe two types of measurements, formative and reflective, which provide the theoretical basis for computing the first step of the PLS-SEM. In the second step, we present the regression model to estimate the path coefficients. For this, we also consider a stepwise variable selection procedure to select the final set of variables related to personal characteristics.

4.3.2. Formative Measurements

A formative construct is defined by specific individual aspects, each captured by an indicator. The construct is formed by its indicators, each representing a different aspect of the latent construct [99]. The six formative constructs used in our model are derived from the literature reviewed in Section 2. They include comfort ($COM$), safety ($SAF$), health ($HEA$), performance expectation ($PEX$), technology affinity ($TAF$), and knowledge and preference ($KAP$). The constructs have been validated in various technology adoption contexts, providing a robust theoretical foundation. In addition, they have been validated on the data ([14] (Table 7)). The following equations define the composite scores of the six formative constructs that appear in our model.

$$\begin{array}{cc}\hfill CO{M}_i& ={\omega }_{BRE}·BR{E}_i+{\omega }_{HIO}·HI{O}_i+{\omega }_{VEN}·VE{N}_i,\hfill \end{array}$$

(1)

$$\begin{array}{cc}\hfill SA{F}_i& ={\omega }_{SOS}·SO{S}_i+{\omega }_{SBO}·SB{O}_i+{\omega }_{RPR}·RP{R}_i,\hfill \end{array}$$

(2)

$$\begin{array}{cc}\hfill HE{A}_i& ={\omega }_{HMA}·HM{A}_i+{\omega }_{HEM}·HE{M}_i+{\omega }_{HEN}·HE{N}_i+{\omega }_{APR}·AP{R}_i,\hfill \end{array}$$

(3)

$$\begin{array}{cc}\hfill PE{X}_i& ={\omega }_{ESI}·ES{I}_i+{\omega }_{HOM}·HO{M}_i+{\omega }_{AMO}·AM{O}_i,\hfill \end{array}$$

(4)

$$\begin{array}{cc}\hfill TA{F}_i& ={\omega }_{TEX}·TE{X}_i+{\omega }_{TPI}·TP{I}_i+{\omega }_{TXT}·TX{T}_i,\hfill \end{array}$$

(5)

$$\begin{array}{cc}\hfill KA{P}_i& ={\omega }_{KLE}·KL{E}_i+{\omega }_{CAP}·CA{P}_i+{\omega }_{HAP}·HA{P}_i.\hfill \end{array}$$

(6)

The weights $\omega$ are the standardized weights of each observed indicator in the formative construct. Therefore, they reflect the contribution of each indicator, considering the relationships between them. Specifically, the weights $\omega$ were standardized to facilitate comparison and interpretation. In PLS-SEM, a formative construct can be expressed as a multiple regression $F={\omega }_1{X}_1+{\omega }_2{X}_2+$...$+{\omega }_n{X}_n+\epsilon ,$ where F is the composite score for the formative construct, $X_1$ through $X_n$ are the indicators, and ${\omega }_1$ through ${\omega }_n$ are the standardized regression weights. In PLS-SEM, there is no intercept coefficient because of data standardization. Additionally, there are generally no error terms for formative measures because using several indicators to measure a construct is considered to capture all aspects [100]. As a result, we drop the error term from the equations. The index i refers to the observations in the data.

4.3.3. Reflective Measurements

A reflective construct assumes that the construct causes all indicators. Thus, the indicators are highly correlated and interchangeable in meaning [99]. Our model contains only one reflective construct, the intention to adopt SH ($AIN$), which is measured by three indicators: intended usage ($IUS$), predicted usage ($PUS$), and opportunistic usage ($OUS$). The variables measure the intention to adopt SH technology. We can ensure higher reliability and validity by utilizing three variables, which were asked and formulated at different points in the survey. The variables and the resulting construct are drawn from previous studies on SH adoption (see, e.g., [75,76,78]). A set of three equations (see Equation (7)) reflects the relationship from the construct to each of the indicators, based on bivariate regressions while accounting for measurement error [92]:

$$\{\begin{array}{cc}\hfill IU{S}_i& ={\iota }_{IUS}·AI{N}_i+{\epsilon }_{IUS,i},\hfill \\ \hfill PU{S}_i& ={\iota }_{PUS}·AI{N}_i+{\epsilon }_{PUS,i},\hfill \\ \hfill OU{S}_i& ={\iota }_{OUS}·AI{N}_i+{\epsilon }_{OUS,i}.\hfill \end{array}$$

(7)

As a result, we obtain correlation weights between the construct and each of its indicators, which PLS-SEM ultimately combines into the final indicator loadings ${\iota }_{IUS}$, ${\iota }_{PUS}$, and ${\iota }_{OUS}$. We assume that the error terms $\epsilon$ are uncorrelated with each other and the construct $AIN$, resulting in a mean value of zero.

4.3.4. Structural Models

In the second step of PLS-SEM, we use OLS regression to estimate the structural models. The hypothesized model comprises two endogenous constructs, performance expectation ($PEX$) and adoption intention ($AIN$), which results in the regression Equations (8) and (9).

Equation (8) shows the regression model depicting the relationship between performance expectation and the three exogenous constructs perceived comfort ($COM$), safety ($SAF$), and health ($HEA$) benefits:

$${\overline{PEX}}_i={\lambda }_0+{\lambda }_{COM}·CO{M}_i+{\lambda }_{SAF}·SA{F}_i+{\lambda }_{HEA}·HE{A}_i+{\epsilon }_{PEX,i}$$

(8)

where we use the notation ${\overline{PEX}}_i$ to distinguish it from the composite score $PE{X}_i$ defined in Equation (4). The $\lambda$s are the estimated intercept and coefficients, respectively, and ${\epsilon }_{PEX}$ is the error term. Since the construct scores that result from the first step represent standardized values, PLS-SEM also applies data standardization in the regression. As a result, the intercept is zero, and the coefficients range between −1 and +1. Furthermore, a coefficient of +1 indicates a strong positive correlation, while negative values indicate a negative correlation. Values close to zero are associated with weaker relationships [92].

Equation (9) describes the regression model that links the principal variable of interest, the adoption intention ($AIN$), with its respective variables. The model in Figure 1 consists of two parts. The variables pointing from the left of the graph to the $AIN$ variable represent different aspects of SH benefits, including comfort ($COM$), safety ($SAF$), health ($HEA$), and performance expectation ($PEX$). They make what we call the “base model”.

The variables pointing from the right of the graph relate to the personal characteristics of a potential SH user. The final set of variables used in the model (see Figure 1) results from a stepwise selection procedure used to understand which personality traits described by the 34 variables available in the survey data (see Section 4.2) contribute significantly to explain the adoption of SH. Starting from the “base model”, we incrementally added explanatory variables to find the model that best describes the intention to adopt. A variable was only kept if the resulting regression model yielded an improved R-squared value, a lower value of the Bayesian information criterion (BIC), significant regression coefficients (p-value below 0.05), and no multicollinearity problems (variance inflation factor, VIF, value below 3.3); see, for example, Arthanat et al. [87]. Details of the intermediate regression models verified during the selection procedure are provided in

Table A2. Performance results for intermediate models in the stepwise variable selection procedure in Regression Model (9).

Round	Model	Extension	Coeff.	Sig.	Adj. ${\mathit{R}}^2$	BIC	Non-Sign. Variable	Inner VIF	Outer VIF
0	Base *		na.	na.	0.386	−700.8	None	OK	OK
1	Base	$TAF$	0.385	0.000	0.505	−1019.7	None	OK	OK
2	Round 1	$KAP$	0.284	0.000	0.543	−1131.3	None	OK	OK
3	Round 2	$GEN$	0.196	0.000	0.551	−1153.1	None	OK	OK
4	Round 3	$AGE$	−0.096	0.000	0.560	−1175.7	None	OK	OK
5	Round 4	$FRA$	−0.119	0.004	0.562	−1176.6	None	OK	OK
6	Round 5	$HOW$	0.112	0.001	0.565	−1179.5	None	OK	OK
7	Round 6	$CAL$	0.061	0.001	0.568	−1184.0	None	OK	OK

Notes: The columns “Coeff”. and “Sig” refer to the regression coefficient and significance (p-value) for the added variable named in the column “Extension”. The column “Non-sig. variable” indicates whether any of the regression coefficients of the variables results in a significance level (p-value) worse than 5% when the extension is added. The last columns (“Inner VIF” and “Outer VIF”) indicate whether any variable exceeds the VIF threshold of 3.3. * The “base model” refers to the model that only includes the variables $COM$, $SAF$, $HEA$, and $PEX$, as described in Section 4.3. The abbreviation “na” stands for “not applicable”.

in the Appendix A. The final regression model includes the constructs of technology affinity ($TAF$) and knowledge and preference ($KAP$), as well as age ($AGE$), gender ($GEN$), frailty ($FRA$), homeownership ($HOW$), and cultural activity level ($CAL$).

The model for the intention to adopt SH writes out as follows:

$$\begin{array}{cc}\hfill AI{N}_i=& {\beta }_0+{\beta }_{COM}·CO{M}_i+{\beta }_{SAF}·SA{F}_i+{\beta }_{HEA}·HE{A}_i+{\beta }_{PEX}·PE{X}_i\hfill \\ \hfill & +{\beta }_{TAF}·TA{F}_i+{\beta }_{KAP}·KA{P}_i+{\beta }_{AGE}·AG{E}_i+{\beta }_{GEN}·GE{N}_i\hfill \\ \hfill & +{\beta }_{HOW}·HO{W}_i+{\beta }_{FRA}·FR{A}_i+{\beta }_{CAL}·CA{L}_i+{\epsilon }_{AIN,i}.\hfill \end{array}$$

(9)

From the standardization applied in step one, the intercept ${\beta }_0$ is zero; the other $\beta$s are the estimated regression coefficients, and ${\epsilon }_{AIN}$ denotes the error term.

5. Results

In this section, we present the results of the PLS-SEM calibration using SmartPLS (version 4) software. First, we validate the measurement models defined using Equations (1)–(7). Then, we provide the results of the regression models for performance expectation (Equation (8)) and adoption intention (Equation (9)). When presenting the results, we discuss key metrics on the performance of the models [101].

5.1. Validation of the Measurement Models

5.1.1. Formative Measurements

To evaluate the validity of each formative construct, we discuss a set of metrics as suggested by Becker et al. [94] and Hair et al. [92]. First, content validity refers to the degree to which the selected indicators accurately and comprehensively capture the content of a construct. The goal is to ensure that the indicators effectively represent the depth of a construct. The survey was based on strong theoretical and empirical foundations following a literature review, exploratory interviews, and pre-tests to achieve this (see the procedure described in [14] (Appendix B)). This approach aligns with established practices in the literature, thereby enhancing construct validity. Previous studies have discussed and validated the constructs that we use, further supporting their application in our model [17,102,103]. Second, indicator collinearity refers to a potential correlation between indicators, which can negatively impact the standard error of the indicator weights and complicate the estimation of each indicator’s unique contribution. We use the VIF metric to assess the collinearity of indicators and find that all indicators’ VIF values are below the cautious threshold of 3.3. We observe that the health maintenance ($HMA$), safety booster ($SBO$), and health monitoring ($HEM$) variables exhibit the highest VIF values of 2.832, 2.795, and 2.625, respectively. Third, we assess the significance and relevance of the indicators. The values of the indicators’ weights and associated significance levels provide information on the relative importance of each indicator in the corresponding construct. In PLS-SEM, the weights $\omega$ are the standardized weights of each observed indicator in the formative construct, reflecting the contribution of each indicator while considering the relationships between them. These weights were standardized to facilitate comparison and interpretation, as discussed in Section 4.3. We report these levels in

Table 4. Validation of the formative and reflective constructs.

Equation	Construct	Indicator		Coefficient	Sig.
Formative constructs
(1)	Comfort	Burden relief	${\omega }_{BRE}$	0.261	***
		Home information	${\omega }_{HIO}$	0.700	***
		Value enhancement	${\omega }_{VEN}$	0.243	***
(2)	Safety	Sense of safety	${\omega }_{SOS}$	0.552	***
		Safety booster	${\omega }_{SBO}$	0.169	**
		Risk protection	${\omega }_{RPR}$	0.419	***
(3)	Health	Health maintenance	${\omega }_{HMA}$	0.191	**
		Health monitoring	${\omega }_{HEM}$	0.496	***
		Health encouragement	${\omega }_{HEN}$	0.284	***
		Accident prevention	${\omega }_{APR}$	0.178	***
(4)	Performanceexpectancy	Everyday simplification	${\omega }_{ESI}$	0.463	***
		Home monitoring	${\omega }_{HOM}$	0.378	***
		Activity motivation	${\omega }_{AMO}$	0.310	***
(5)	Technologyaffinity	Technology experimenter	${\omega }_{TEX}$	0.642	***
		Technology pioneer	${\omega }_{TPI}$	0.407	***
		Technology expert	${\omega }_{TXT}$	0.058
(6)	Knowledge andpreference	Knowledge level	${\omega }_{KLE}$	0.453	***
		Convenience application	${\omega }_{CAP}$	0.524	***
		Health application	${\omega }_{HAP}$	0.387	***
				Loading	Cronbach’s $\alpha$
Reflective construct
(7)	Adoptionintention	Intended usage	${\iota }_{IUS}$	0.965
		Predicted usage	${\iota }_{PUS}$	0.970	0.960
		Opportunistic usage	${\iota }_{OUS}$	0.952

Note: The significance values in column “Sig” are coded as follows: ** p < 0.01, *** p < 0.001.

. We observe that, through all constructs, the coefficients of the indicators yield high significance levels, with a single exception that the indicator related to technology expertise (${\omega }_{TXT}$) is not significant in the technology affinity construct (Equation (5)). However, we retain it as an indicator for content validity reasons. The literature on SH users’ characteristics suggests that measuring technology affinity should include several aspects related to technology, such as ownership, expertise, and familiarity (see Section 2.2). With the above, we conclude that all constructs are valid.

5.1.2. Reflective Measurements

We evaluate reflective measurements for their reliability and validity using metrics related to indicator reliability, internal consistency, convergence validity, and discriminant validity [92,99]. The reflective measurement adoption intention ($AIN$) meets the commonly accepted thresholds for the metrics and is thus reliable and valid for our research context. First, indicator reliability is demonstrated by the squared values of the loadings, which represent the strength and direction of the relationships between the latent construct and its observed indicators (see Table 4). We find the following squared values for the three indicators: intended usage ($IUS$) 0.931, predicted usage ($PUS$) 0.941, and opportunistic usage ($OUS$) 0.906. All values exceed the commonly accepted threshold of 0.708 (corresponding to the construct explaining at least 50% of the indicator’s variance). Second, internal consistency assesses the ability of indicators to measure the same underlying construct. It is typically evaluated using Cronbach’s alpha coefficient (threshold $>0.6$) and the composite reliability (threshold $>0.6$). Both measures demonstrate satisfactory values, with a Cronbach’s alpha of 0.960 and composite reliability of 0.974. Some researchers mention upper thresholds of 0.95 for both metrics, which could suggest redundant items. We have adopted the construct of adoption intention ($AIN$), along with its three indicators from previous studies that do not indicate any limitations (see, e.g., [75,76,78,102]). Regarding convergence validity in reflective measurements, the correlation between a measure and a comparable measure of the same construct is assessed through the average variance extracted (AVE). A value greater than 0.5 is generally considered satisfactory, indicating that the construct explains approximately 50% of the variance in its indicators. We find a high value of 0.926. Discriminant validity refers to the degree to which a construct is genuinely distinct from other constructs, indicating the uniqueness of the construct within the model. The Heterotrait–Monotrait (HTMT) ratio measures this metric. Our model comprises only one reflective construct, so we could not determine the discriminant validity using the conventional Heterotrait–Monotrait ratio. At the bottom of Table 4, we report the loading on the intention to adopt SH ($AIN$) construct.

5.2. Results for the Structural Models

Before reporting the results of the regression models in Equations (8) and (9), we provide information on the results’ validity and the models’ explanatory power. Regarding collinearity, we observe that all VIF values of the inner model are below the cautious threshold of 3.3, indicating no collinearity problems. Looking into Model (8), the comfort construct ($COM$) exhibits the highest VIF value with 2.267. In Model (9), the performance expectation ($PEX$) construct has the highest VIF value of 3.062. These findings show that our model meets the collinearity thresholds suggested for PLS-SEM analyses. Considering explanatory power, Model (9), which represents our principle variable of interest, has a satisfactory R-squared value of 0.571 and an adjusted R-squared value of 0.568, comparable to similar studies on SH adoption. Such studies generally fall into three categories. The first category employs a validated technology acceptance framework and shows R-squared values ranging from 0.610 [75] to 0.820 [78]. The second group comprises studies that rely on validated frameworks but adapt them to address specific research questions. Our research design is closest to these studies that report R-squared values ranging from 0.310 [104] to 0.540 [81]. The third group includes studies that develop their model. Here, R-squared values range from 0.080 [87] to 0.426 [31]. We note that including personal characteristics variables significantly enhances explanatory power, increasing the adjusted R-squared value from 0.386 to 0.568. This inclusion also improves the BIC value from −700.8 to −1184.0. Note that we calculated BIC values for different model variations to control for overfitting. Additionally, we observe that Model (8) exhibits an R-squared value of 0.640 and an associated BIC value of −1508.0.

5.2.1. Regression Model (8) for Performance Expectation

In

Table 5. Results of the regression model (8) for performance expectation.

Variable		Coefficient	Sig.	Hypothesis
Prevention
Safety $SAF$	${\lambda }_{SAF}$	0.295	***	(H1)
Health $HEA$	${\lambda }_{HEA}$	0.229	***	”
Comfort
Comfort $COM$	${\lambda }_{COM}$	0.389	***	(H3)

Note: See Table 4. *** p < 0.001.

, we report the results of the regression model in Equation (8). It shows how perceived benefits related to comfort, safety, and health benefits are associated with performance expectations. The three benefits are significant for the overall perceived usefulness of the technology. Comfort considerations seem most relevant, followed by safety and health benefits.

5.2.2. Regression Model (9) for Adoption Intention

In

Table 6. Results of the regression model (9) for adoption intention.

Variable		Coefficient	Sig.	Hypothesis
Prevention
Safety $SAF$	${\beta }_{SAF}$	0.070	*	(H2)
Health $HEA$	${\beta }_{HEA}$	0.068	**	”
Comfort
Comfort $COM$	${\beta }_{COM}$	0.107	***	(H4)
Performance expectation
Performance expectation $PEX$	${\beta }_{PEX}$	0.110	***	(H5)
Personal characteristics
Technology affinity $TAF$	${\beta }_{TAF}$	0.273	***	(H6)
Knowledge and preference $KAP$	${\beta }_{KAP}$	0.245	***	”
Age $AGE$	${\beta }_{AGE}$	−0.108	***	”
Gender $GEN$ (baseline: female)	${\beta }_{GEN}$	0.218	***	”
Frailty $FRA$ (baseline: no)	${\beta }_{FRA}$	−0.151	***	”
Homeownership $HOW$ (baseline: renter)	${\beta }_{HOW}$	0.116	**	”
Cultural activity level $CAL$ (baseline: hardly ever)	${\beta }_{CAL}$	0.061	**	”

Note: The significance values in column “Sig” are coded as follows: * p < 0.05, ** p < 0.01, *** p < 0.001.

, we present the results of the regression model in Equation (9), which demonstrates the association of the considered variables to the intention to adopt SH. Preventive health and safety benefits and comfort considerations are significantly associated with a greater intention to adopt. All three coefficients attain a significance level of 5%. Regarding personal characteristics, we observe that traits related to technology exhibit the strongest relationship with adoption intention. When combined with the user’s gender, these features produce regression coefficients that, in absolute terms, exceed all the ones of the three benefits, highlighting their significance. Although the current literature focuses mainly on age as the primary attribute of an SH user, our findings show that the impact of frailty, homeownership, and cultural activity level is on par with that of age.

6. Discussion

Our study contributes to the risk and insurance literature by providing insights into users’ perceptions of risk prevention benefits in the context of SH technology. Whereas wearable and telematic devices have received considerable research attention, SH technology has yet received less attention. Considering the potential for this technology to optimize risk behavior and enable real-time management, and with an increasing number of households using SH devices, the attention of risk researchers and practitioners is warranted. This chapter is divided into three parts: the discussion of results, theoretical implications, and practical implications.

6.1. Discussion of Results

In Figure 2, the summarized results demonstrate the significance levels and regression coefficients, illustrating the strength and direction of variables’ relationships with the intention to adopt SH.

Preventive health and safety benefits are significantly connected to higher performance expectancy and a higher intention to adopt SH. This supports both hypotheses (H1) and (H2), which posit that safety and health prevention benefits correlate positively with performance expectations and SH adoption intention. The data shows particularly pronounced significant levels ($p<0.001$) between these variables and performance expectancy, indicating that users easily associate benefits of SH technology to safety and health areas, or see SH as a viable tool to improve home safety and health. Similarly, hypotheses (H3) and (H4), which state that comfort benefits correlate positively with both performance expectations and SH adoption intention, are strongly confirmed by the data (for both, $p<0.001$). Comfort seems to be a primary driver for users, as it has a greater impact than safety and health benefits, emphasizing its crucial role in the adoption decision and its significant impact on performance expectations. Moreover, performance expectations positively correlate with adoption intention, supporting hypothesis (H5). While the data confirm a pronounced positive relationship ($p<0.001$), the strength of the relationship is comparable to the one of comfort benefits. Additional personal characteristics related to technology, sociodemographics, and active aging lifestyle also correlate significantly with adoption intention, confirming hypothesis (H6). Where the strongest positive effects are observed for technology-related variables, age and frailty show negative effects. The integrated view of user expectations on risk prevention and personal characteristics provides an understanding of the factors influencing SH adoption, and these insights can inform strategies to promote SH adoption and enhance user engagement.

6.2. Theoretical Implications

While previous research has shown that performance and convenience expectations positively influence the adoption of SH technology, we shed light on the extent to which risk prevention benefits play a role. This section discusses our results against the theoretical background.

Delving further into the importance of safety and health, our empirical results indicate that prospective SH users perceive technology as a viable tool to improve home safety and health. Safety-related SH products, such as fire alarms, security cameras, and water sensors, are commonly known to the public [105]. The literature also highlights some prevention use cases related to health [106]. While prior studies have often focused on individual diseases [87] and show rather low levels of adoption [31], our study focus on health prevention benefits before the manifestation of a disease. With that, we find significant effects, which may be affected by the age of 45 years and older in our respondents—consistent with other research [68]. The importance of comfort benefits concerning SH adoption is critical and consistent with previous research findings [1,14,75]. The work of Chang and Nam [12] conducted in the Republic of Korea revealed similar patterns in the relative importance of comfort, health, and safety, with energy considerations found to be insignificant, implying that cultural differences may have a minor influence on our research results. Finally, comparing the personality traits that characterize potential SH users in our study, we find previously unstudied variables like frailty, homeownership, and cultural activity level. These exert similar influences on adoption intention as, for example, the widely studied variable age [76]. The integration of these personality traits also significantly improves the explanatory power of the model, emphasizing the importance of tailored market segmentation for SH users with a focus on prevention-related features.

6.3. Practical Implications

Practitioners and policymakers may find the results helpful. In particular, insurers interested in leveraging SH technology for innovative technology-driven service models can benefit from the established evidence. First, it is essential to contextualize the value proposition for safety and health benefits within the broader range of user comfort and performance expectations. Focusing solely on preventative aspects may have an adverse effect when users perceive the technology as overbearing or patronizing. Regarding market segmentation, the data suggest that homeowners and active individuals represent promising target groups. Marketing strategies for prevention-oriented SH solutions could thus be tailored accordingly. Linking SH solutions with an active lifestyle could be a promising approach. Thirdly, the findings indicate that a lack of knowledge and affinity for technology could pose significant obstacles to dissemination, thus necessitating the provision of offers such as concierge or support services to mitigate such barriers.

Generally, user behavior is rarely emphasized or discussed in the context of SH. However, to fully realize the potential of SH technology for risk prevention, it may be necessary to encourage user engagement in SH prevention use cases. Such engagement can pave the way for insurers seeking to minimize risk and control claims costs. Working collaboratively with individuals to proactively implement risk-mitigation measures through SH allows insurers to actively promote safety and health measures. Insurers can strengthen their relationship with policyholders by providing incentives for household prevention efforts, such as premium discounts or financial contributions towards acquiring and maintaining technology. Maintaining user engagement and compliance with the intended service (e.g., keeping the SH devices turned on and changing batteries) is also crucial for the effective functioning of the technology and the management of emerging cybersecurity risks.

7. Conclusions

This work investigated prevention benefits in SH adoption and empirically derived the extent to which users perceive the value of prevention in technology. We identified and established clear links between the prevention benefits associated with safety and health and the intention to adopt SH technology. Furthermore, we confirmed a significant and positive relationship between the comfort benefits of SH and the increased interest in the technology. We also provide a set of personal characteristic variables that describe the individuals attracted to the SH prevention premise.

Although we advance our understanding of SH’s potential in terms of prevention, inherent limitations due to the self-reported nature, limited knowledge of some respondents, a lack of time dimension, and the geographic focus of the data set limit the scope and generalizability of our findings. Future research could benefit from controlled experiments or observations of purchase and usage behavior to bridge the gap between stated adoption intentions and actual smart home adoption and risk prevention behaviors. In addition, exploring its role and integration into the insurance sector, as well as potential adaptations to established technology acceptance models, offer promising directions for further investigation.