Identifying and validating housing adaptation client profiles – a mixed methods study

Abstract

Purpose: An increasing number of people will live with disabilities in their homes and consequently, the need for home-based interventions will increase. Housing adaptations (HAs) are modifications to the physical home environment with the purpose to enhance independence for a heterogeneous group of people. Increasing the knowledge of the characteristics of HA clients by exploring their heterogeneity, could facilitate the planning of interventions and allocation of resources. The purpose of this article was to identify and validate HA client profiles.

Materials and methods: This cross-sectional study applied a mixed methods design to identify profiles of HA clients through cluster analysis confirmed by qualitative interview data. The sample consists of 241 HA clients in Sweden with a mean age of 75.1 years.

Results: A classification into five groups emerged as the one best describing the heterogeneity of characteristics among this sample of clients. Five client profiles were outlined based on their age and level of disability, and the variation between the profiles was confirmed through the qualitative interview data.

Conclusions: The identified client profiles are a step towards a better understanding of how home-based interventions could be delivered more effectively to groups of HA clients, based on their different characteristics.

Implications for rehabilitation

• Housing adaptations are structural modifications to the physical home environment with the purpose to enhance independence for people with disabilities.

• People applying for housing adaptations are a heterogeneous group with different needs.

• This study outlines five client profiles which can guide professionals on how to differentiate home-based interventions and follow-up processes among housing adaptation clients.

Keywords:

• Home modifications
• heterogeneity
• disability
• cluster analysis

Introduction

In Sweden, an aging-in-place policy aims for people to be able to continue living in their home despite disease and functional decline [1]. This will assumingly result in an increasing number of people living with disabilities in their own home since disability is more prevalent in the fast-growing older age groups [2]. Consequently, the need for home-based interventions within health care and rehabilitation is likely to increase as well. Home-based interventions are generally considered as complex, as described by the UK Medical Research Council, since they consist of several interacting components and are difficult to standardize [3]. As an additional challenge, these complex interventions are often delivered to populations with great individual variation regarding age, health conditions, and functional abilities [4, 5].

A housing adaptation (HA), defined as structural modifications to the physical home environment with the purpose of enhancing independence [6], is an example of a complex intervention delivered to a heterogeneous population. According to Swedish legislation persons with functional impairments are entitled to a grant covering the full cost of an HA [7]. The need for the intervention must be assessed and verified by a health professional, most often an occupational therapist, and an HA application subsequently submitted to the municipality [6]. In Sweden, with a population of approximately 10 million people [8], 73 200 HAs were granted in 2015 at a total cost of 1039 million Swedish kronor (approx. 102 million euro). The most commonly performed interventions were adjustment of thresholds, mounting of grab bars, installation of ramps, stove guards and adaptation of bathrooms [9].

Previous research indicates that HAs have a positive effect on outcomes such as activity [10–14], usability of the home [10, 15] and falls [16, 17]. HAs thereby have the potential to counteract disability by targeting the environmental factor within this multi-dimensional concept as it is defined by the World Health Organization (WHO) in the International Classification of Functioning, Disability and Health (ICF) [18]. According to the ICF, functioning and disability are dynamic interactions between health conditions and contextual factors; both personal and environmental. Functioning and disability encompass three components: body functions, activity and participation. Activity is defined as the execution of a task and participation as the involvement in a life situation. Disability should be viewed as an umbrella term for physical impairments, activity limitations and participation restrictions, which are the negative aspects of the interaction between a person with a health condition, and that person’s contextual factors [2].

Thus, there is supportive evidence regarding positive effects of HAs on disability-related outcomes. However, the effects of HAs for the individual clients are not being evaluated according to any mutual standardized methods among Swedish municipalities [19]. An on-going project has addressed this lack of evidence-based guidelines and is currently evaluating a structured, research-based strategy for managing HAs in three municipalities in southern Sweden [20]. As part of this project, the importance of exploring the population of HA clients has been pointed out as well [21]. Increasing the knowledge of their characteristics could facilitate the implementation of structured client assessments and follow-ups [4, 22, 23]. Furthermore, learning more about differences between groups of HA clients could contribute to assuring that resources allocated to the intervention benefit those in most need, and that clients in most need of follow-ups afterwards are contacted.

Available information about HA clients primarily comes from statistical reports regarding age, gender and cost. In Sweden in 2015, 72% of the clients were older adults over the age of 70; 60% were women and 40% men [9]. A recent study within the aforementioned project explored the heterogeneity of HA clients, employing hierarchical cluster analysis, and identified six groups of participants with different characteristics who were found to differ significantly regarding participation and self-rated health [21]. The results of the clustering were regarded as preliminary; however, since only approximately half of the final study sample was available at the time. A larger number of participants could be expected to provide more representative results, which is why it was considered of great relevance to continue exploring the heterogeneity of HA clients including the entire study sample. In order to strengthen the validity of the findings and to further increase the understanding of the population, available qualitative data were used as well in the current study. Thus, the purpose of this study was to identify and validate HA client profiles.

Materials and methods

Design and setting

This is a cross-sectional study employing a triangulation mixed-methods design [24]. Quantitative and qualitative baseline data originally collected for the Research Strategy for Housing Adaptations (ResHA) study, a larger longitudinal study of HA clients in Sweden [20], was extracted from a survey database as well as from transcribed interviews. Quantitative and qualitative data were analyzed separately and then converged during the interpretation. This convergence model, a variant of the triangulation design, can be used to “validate, confirm or corroborate quantitative results with qualitative findings” [24].

Sampling and participants

The participants, who were all applicants of HA grants, were identified and included for the ResHA study [20] by the municipality occupational therapists, who conduct HA needs assessments. Individuals at least 20 years of age applying for an HA grant were eligible for inclusion. Applicants living in sheltered housing and/or not being able to communicate or follow instructions in Swedish were excluded. Eventually, 241 participants were included between January 2013 and December 2015; 63.1% were women and the mean age of the sample was 75.1 years. Additional demographic information about the sample is presented in Table 1.

Table 1. Characteristics of the sample (n= 241).

Age, mean (SD) (n= 236)	75.1 (14.0)
Gender, no. (%)
Female	152 (63.1)
Male	89 (36.9)
Education in years, mean (SD) (n= 230)	9.7 (3.2)
Area lived in, no. (%)
Larger urban area	155 (64.3)
Smaller urban area	52 (21.6)
Rural area	33 (13.7)
Missing	1 (0.4)
Dwelling, no. (%)
Apartment	131 (54.4)
One-family house	97 (40.2)
Other	4 (1.7)
Missing	9 (3.7)
Social situation, no. (%)
Living alone	134 (55.6)
Living with someone	106 (44.0)
Missing	1 (0.4)
Use of mobility devices indoor, no. (%)
Yes	191 (79.3)
No	50 (20.7)
Use of mobility devices outdoor, no. (%)
Yes	227 (94.2)
No	14 (5.8)

Out of the 241 participants, 17 individuals were purposively selected for in-depth interviews. They were selected with the intention to include a diverse range of participants in terms of gender, age, civil state, level of ADL dependence, use of mobility aids, type of housing and type of HA applied for [25].

Data collection

Collection of quantitative data was conducted by occupational therapists, trained in the data collection methodology, during home visits through observations and the clients’ self-assessments. Validated instruments [15, 26–29] as well as study-specific questions [20] were used to explore several aspects of housing and health, listed below.

• Dependence in ADL assessed with the ADL staircase [29]. The instrument consists of nine items on activities of daily living; feeding, transferring, using the toilet, dressing, bathing, cooking, transportation, shopping, and cleaning. They were measured on a four-point scale (0–3) as “independent without difficulty”, “independent with difficulty”, “partly dependent,” and “dependent”, and added up to a total score (0–27) for dependence.

• Number of functional limitations (0–12) measured as present or absent using 12 items from the personal component in the Housing Enabler Instrument [26]. The higher the score the larger the number of functional limitations. Included items were for example limitations of vision, balance, coordination, upper and lower limb function, and dependence on a walking aid or wheelchair.

• Cognitive impairment assessed with the Montreal Cognitive Assessment (MoCA), which measures different aspects of cognitive function on a 0–30 scale. A score of 25 points or below is considered to indicate cognitive impairment [28].

• Concerns about falling measured with the short version of the Falls Efficacy Scale-International, FES-I. The instrument includes seven activities, each assessed on a four-point scale (1–4), adding up to a total score (7–28) with a higher score indicating more fear [27].

• Usability of the home assessed with the Usability In My Home (UIMH) instrument, which is a self-rating tool measuring satisfaction when performing different activities in the home environment on a 1–5 scale, with a higher score indicating a higher level of satisfaction [15]. A score of 0 was attributed to those who are not performing the [21] activity, as suggested by an ongoing validation work by Malmgren Fänge et al. [30]. Three aspects of usability were included: self-care, social and leisure/outdoor. The self-care aspect comprising five items (going to the toilet, personal hygiene, cooking, preparing snacks and moving around the home with or without mobility device), the social aspect with three items (socializing with family and friends in the home, contacting other people using the telephone or Skype and watching TV or listening to the radio) and finally, the leisure and outdoor aspect incorporating three items (entering the house, picking up mail and engaging in leisure activities in the home). Total scores were consequently 0–25 for self-care and 0–15 for social and leisure/outdoor aspects, respectively [21].

• Study specific questions were demographic data, participants’ social situation such as living alone or with someone (could be a spouse, friend, lodger, child, or parent); kind of dwelling or area lived in (larger urban area was defined as a city; smaller urban area was defined as an area with several houses and with access to a supermarket or other services; rural area was defined as the countryside with no access to services nearby).

Qualitative data were gathered through in-depth interviews with purposively sampled participants. The 17 participants were interviewed in their homes by a researcher and occupational therapist, experienced at conducting qualitative interviews. An interview guide was used, consisting of open-ended questions formulated as follows:

• Would you like to describe what made you come to apply for a HA?

• Can you describe the problems you have? You are welcome to show me in what situations you experience those problems.

• How long have you had this need or problem?

• How come you applied at this time?

• Have you tried some other solutions or strategies to overcome the problems

• you describe? If so, what kind of solutions or strategies?

• Can you describe your experience of the process of applying for a HA?

• What are your expectations of what the housing adaption will mean for you in your everyday life?

The interviews were recorded and subsequently transcribed verbatim [25].

Data analysis – quantitative

Cluster analysis was used to identify meaningful structures among the HA clients.

This is an exploratory method for organizing the information in a dataset to make it easier to understand. Groups are created based on patterns of similarities and differences, so that individuals within a group – or cluster – have more in common with each other than with those in other clusters. A classification of individuals through cluster analysis does not provide a result that can be considered in terms of “true” or “false”, but should rather be judged by its usefulness [31].

The variables included in the cluster analysis in this study, to identify groups of HA clients with similar characteristics, were age, dependence in ADL, functional limitations, cognitive impairment, concerns about falling and usability of the home. They were selected following the theoretical model proposed by Thordardottir et al. [21]. The variables with minimum, maximum, and mean values for the sample are presented in Table 2.

Table 2. Variables included in the cluster analysis (total sample n= 241).

Variable	Min–max in the sample	Mean (SD) in the sample
Age (n= 236)	25–95	75.1 (14.0)
ADL dependence^a (n= 233)	0–25	11.6 (5.7)
No. of functional limitations^b (n= 241)	1–11	4.8 (1.9)
Cognitive impairment^c (n = 184)	10–30	22.7 (4.6)
Concerns about falling^d (n= 216)	6–28	15.4 (5.7)
Usability in self-care^e (n= 241)	0–25	17.4 (5.9)
Usability in social relations (n= 241)	0–15	11.7 (3.4)
Usability in leisure/outdoors (n= 241)	0–15	8.0 (3.9)

a Assessed with the Activities of Daily Living (ADL) staircase.

b Assessed with 12 items from the personal component of the Housing Enabler Instrument.

c Assessed with the Montreal Cognitive Assessment (MoCA).

d Assessed with the short version of the Falls Efficacy Scale International (FES-I).

e Assessed with the Usability in My Home (UIMH) instrument.

SPSS Statistics 24.0 (IBM Corporation, Armonk, NY) was used to perform the analysis, and out of several available cluster methods hierarchical cluster analysis was chosen, employing Ward’s method criterion, squared Euclidean distances and standardized z-scores. Missing values were replaced using average replacement, i.e., substituting the missing value with the mean value of the variable distribution [32], which enabled the inclusion of the entire sample of 241 participants in the analysis. The rationale for this choice, instead of applying list-wise deletion of cases with missing values, was avoiding the loss of information associated with missing values while reducing the influence of the imputed variable on the clustering algorithm result.

Arriving at the optimal cluster solution entails deciding how many clusters best describe the dataset. This decision must to a large extent be made based on informal and subjective criteria [31]. In this analysis, solutions ranging from three to six groups were considered. The results for the solutions were saved in SPSS, and tables were created to make it possible to compare the different solutions to each other.

Data analysis – qualitative

The analysis of the qualitative data began when the cluster analysis was completed, and it was confirmed that there was at least one interviewed participant in each cluster. Employing content analysis [33], all 17 interviews were first read in their entirety to obtain an overview of the material. Then, each interview was read again, this time focusing on the part of the interview where the participant mainly answered the first two questions of the interview guide. For each interview separately, meaning units [33] were identified and assigned to five predetermined categories corresponding to the variables included in the cluster analysis: ADL dependence, functional limitations, cognitive impairment, concerns about falling and usability of the home. The extracted meaning units were used to construct a narrative describing the participant. The narratives were subsequently compared and contrasted to the quantitative cluster characteristics and client profiles [24].

Ethics

This study was approved by the Ethical Review Board at Lund University 2012/566 and 2013/592. The study was conducted in accordance with the Declaration of Helsinki [34]. Participation in the study was voluntary and all subjects gave their written informed consent for inclusion before participating. All data were collected in the home of the participants, requiring considerable sensitivity to the situation. The population is known to include a considerable number of individuals who could be described as frail and who might have at least moderate cognitive impairment [21]. This should be kept in mind when subjecting this group to multiple assessments and interviews during lengthy home visits. All occupational therapists collecting data had extensive experiences from home visits to people in the participant group, and they were also specifically trained in collecting research data in this context [20]. All data are stored digitally in unidentified and encrypted format, and paper versions are stored in areas accessible only for the researchers in the project.

Results

When comparing the different cluster solutions to each other, it was concluded that important information regarding the sample was lost in the three- and four-group solutions whereas the sixth cluster in the six-cluster solution merely contained a small number of outliers. Thus, the five-cluster solution was regarded as the superior one, providing the most representative description of the sample, based on the included variables. The variable means for each cluster are displayed in Table 3.

Table 3. Characteristics of the five identified profiles (mean ± SD).

Characteristics	1. Older adults with low level of disability (102)	2. Older adults with medium/high level of disability (85)	3. Adults with low level of disability (20)	4. Adults with high level of disability (15)	5. Older adults with medium level of disability incl. at least moderate cognitive impairment (19)
Age	79.5 ± 8.2	79.7 ± 8.4	46.9 ± 8.4	52.1 ± 13.8	79.4 ± 8.2
ADL dependence	8.9 ± 5.0	14.1 ± 4.5	8.0 ± 3.3	18.9 ± 4.7	12.5 ± 5.4
Functional limitations	3.9 ± 1.7	5.2 ± 1.7	4.3 ± 1.7	7.3 ± 1.2	6.0 ± 2.0
Cognitive impairment	24.1 ± 2.7	22.4 ± 3.1	26.7 ± 2.1	22.3 ± 1.7	13.9 ± 3.0
Concerns about falling	13.5 ± 4.2	16.2 ± 5.5	16.9 ± 6.0	20.5 ± 6.9	16.0 ± 4.8
Usability in self-care	21.1 ± 3.4	14.7 ± 5.8	18.2 ± 4.4	8.4 ± 4.5	16.2 ± 5.5
Usability in social relations	13.5 ± 1.5	9.7 ± 3.9	13.3 ± 2.2	10.2 ± 4.2	10.8 ± 2.9
Usability in leisure/outdoors	10.0 ± 3.2	5.3 ± 3.1	8.9 ± 2.5	5.5 ± 3.7	10.5 ± 3.5

The concept of disability was chosen as a theoretical framework to describe and compare the clusters [18]. The clusters were labeled based on age and level of disability, outlining five client profiles, which are described below drawing on the quantitative cluster characteristics as well as the results of the analysis of the interviews. Qualitative data are presented as abbreviated narratives or summaries of narratives when several interviewed participants were classified to the same cluster.

Cluster 1 – older adults with low level of disability (n = 102)

Cluster 1 was characterized by old age, low level of ADL dependence, low number of functional limitations, no/mild cognitive impairment, low level of concerns about falling and usability of the home rated high for all three aspects (self-care, social relations, leisure/outdoor).

The interviewed participants who were classified to this cluster were all older, mostly active and independent, receiving help mainly with instrumental ADL tasks such as house cleaning, laundry and shopping. A personal activity which did present a challenge to some of the participants was taking a shower, and their HA applications concerned adaptations of the bathroom, such as installing a grab-bar or replacing the bathtub with a walk-in shower, to be able to continue showering safely and independently. One client found it increasingly difficult to shower using her tub bench and it was eventually approved that she needed a walk-in shower to enable her to stay independent.

Well anyway, it has gotten worse because my legs have become so heavy and more difficult to lift so… they were here three times to see if I could do it… and then the third time it was approved that I… that it was difficult for me. (Age 91, dependence in ADL 9)

The members of cluster 1 were able to move around inside their home, some needing a mobility device and/or hand rails to be able to do so independently. All of them mentioned going out to do their shopping and/or participating in a social or physical activity such as playing cards with friends, gardening or swimming. They could all go out on their own, although stairs represented a barrier and one person had recently moved to a different apartment building to have access to an elevator while another was in the process of applying for a stair lift. Heavy front or garage doors were other environmental barriers which for a couple of members had led to an HA application to have an automatic door-opener installed.

It will be so much easier, right – to open the garage door. I also open it to bring out some tools and things, so it’s not only for the car. I have some gardening tools and things hanging on a wall. (Age 75, dependence in ADL 8)

Cluster 2 – older adults with medium/high level of disability (n = 85)

Cluster 2 was characterized by old age, medium/high level of ADL dependence, medium number of functional limitations, mild cognitive impairment, medium level of concerns about falling and usability of the home rated medium for self-care and social aspects and low for leisure/outdoor aspects.

The interviewed participants who were members of this cluster required help with both instrumental ADL and some personal ADL tasks. They managed to move around sufficiently inside their homes, although with some effort and/or need of assistance for a couple of them. Since all of them either used a wheelchair or a walker inside, their thresholds had already been removed or they were presently applying to have it done.

The members of this cluster were not able to go outside on their own, however, requiring assistance climbing stairs, opening doors, pushing a wheelchair, etc. For one person, it meant not being able to socialize with the neighbors like she used to, since she had become dependent on her husband to make it up and down the steps.

And the same when I go upstairs. Because when I go up, it’s almost as if… He has to push me a little from behind. I don’t have the strength in my legs, sort of. (Age 78, dependence in ADL 15)

She had applied to have a hand rail installed, hoping that would enable her to manage the steps on her own. Another client expressed great frustration over feeling like a prisoner in his own home since he was not able to run the newly installed stair lift himself, like he had anticipated. A third person felt worried and uncomfortable when going outside since he was dependent on portable ramps which were difficult for the personnel from home services to manage.

…it could happen, if you don’t put them out properly, that you slide off… I’m a little scared that I won’t make it down the ramp since I can’t see anything. (Age 90, dependence in ADL 15)

Cluster 3 – adults with low level of disability (n = 20)

Cluster 3 was characterized by young age, low level of ADL dependence, low number of functional limitations, mild/no cognitive impairment, medium level of concerns about falling and usability of the home rated high for self-care and social aspects and low/medium for leisure/outdoor aspects.

Only one of the interviewed participants belonged to cluster 3. He was experiencing cognitive problems and had applied for assistive technology to enhance his independence and safety at home.

I: Has it happened more than once that you have left the stove and oven on?

P: Yes… yes.

I: And how do you feel about that?

P: Horrible… When you forget things. (Age 41, dependence in ADL 5)

Until recently, he had managed everything on his own but had now decided to also seek assistance with shopping, laundry, and house-cleaning. He had no mobility problems and exercised away from home regularly.

Cluster 4 – adults with high level of disability (n = 15)

Cluster 4 was characterized by young age, high level of ADL dependence, high number of functional limitations, mild cognitive impairment, high level of concerns about falling and usability of the home rated low for self-care, medium for social aspects and low for leisure/outdoor aspects.

The interviewed participants who belonged to this cluster were in need of quite extensive assistance with personal ADL. They required assistance and/or ramps to be able to go outside and an additional common denominator was difficulties moving around inside their homes as well. Some used wheelchairs inside and the barriers were for example thresholds, a narrow doorway to the bathroom or a bathroom too small to receive assistance from another person. For two of the participants in this cluster transfers to and from the toilet or shower were associated with a considerable fall risk. One of them had applied for an HA grant to build an extension to his house making space for a larger bathroom. Both the shower and the toilet were too small for him to be able to transfer safely as well as to receive help with his personal care.

…the toilet is too small. If I have guests downstairs, it’s not so much fun to sit in the hall… and you have to roll out and they have to dry you… Then it’s not so much fun to have guests. (Age 51, dependence in ADL 23)

The other client expressed great frustration over how the thresholds obstructed her daily life. She could not roll her wheelchair into the bathroom by herself and had been putting herself at risk for some time trying to take a few steps leaning on a walker.

And since it can come so suddenly…the leg giving away. To me it can feel ok when I get up and take the walker. But when I start to walk, I feel that it just disappears. Perhaps I make it to the toilet, but then I don’t dare to walk back because I have no strength in my leg. (Age 49, dependence in ADL 18)

Cluster 5 – older adults with medium level of disability including at least moderate cognitive impairment (n = 19)

Cluster 5 was characterized by old age, medium level of ADL dependence, medium/high number of functional limitations, moderate/severe level of cognitive impairment, medium level of concerns about falling, and usability of the home rated medium for all three aspects (self-care, social and leisure/outdoor aspects).

Only one of the interviewed participants was classified to cluster 5. Her ADL dependence varied over time depending on her health status. At times she required hands-on assistance with transfers and personal care. The bathroom was located upstairs and due to difficulties climbing stairs she had applied to have a small toilet built on the ground floor.

…because I have such difficulties walking and have started to wear a diaper on some occasions… I don’t make it upstairs on time. (Age 70, dependence in ADL 12)

She could presently manage without a mobility device inside the house, but had to hold on to walls, furniture, etc. for support. Outside she used walking poles. A walker had been suggested to her, but she did not feel ready for that yet. The assessment of her cognitive function indicated moderate cognitive impairment.

Discussion

It was concluded that a classification into five groups best described the heterogeneity of characteristics among this sample of Swedish HA clients. Five client profiles were outlined based on their age and level of disability [18], and this delineation of five client profiles was corroborated by the qualitative data used in this study.

When labeling the profiles, the aim was to formulate descriptions, which would reflect how their characteristics differed based on the variables included in the cluster analysis. The three older groups were quite homogeneous in terms of their mean age (79.5; 79.7; 79.4) as were the two younger groups with a mean age of 46.9 and 52.1, respectively. Given this classification into three older and two younger groups, the profiles were simply labeled as either “older adults” or “adults”. Furthermore, levels of disability were chosen to distinguish between the profiles since the other variables – ADL dependence, functional limitations, cognitive impairment, concerns about falling and usability of the home – could be linked, directly or indirectly, to various aspects of functioning and disability.

It is important to point out that the levels of disability – low, medium, and high – used here, are relative and do not refer to any specific cutoff values. They are based on comparison between the profiles of this classification and are constructed merely to illustrate and describe their differences. Consequently, the more dependent in ADL, the more functionally limited, the more cognitively impaired, the more concerned about falling and the less satisfied with the usability of the home, in comparison with the other profiles – the higher the level of disability. A majority of the participants were classified to the three groups consisting of older adults – clusters 1, 2, and 5 – which is consistent with available statistics reporting that slightly over 70% of HA clients in Sweden are above the age of 70 [9]. Clusters 1 and 2, characterized by a low and medium/high level of disability respectively, were considerably larger than the other three groups. Although still well-functioning, the older adults of cluster 1 gave an overall impression of being quite vulnerable and at definite risk of declining physical function without for example the automatic door-opener or stair lift they had applied for. Considering the positive relationship between housing and functional ability among older adults [14], HAs might be particularly effective for this group in terms of preventing functional decline [21].

The third, much smaller, group of older adults – cluster 5 – with a medium level of disability was quite similar to cluster 2 regarding the quantitative characteristics, except for moderate/severe cognitive impairment. Since it is common for older adults to continue living in their own home despite disabilities and given the fact that cognitive impairment increases with age [35], it was not surprising that older persons with moderate/severe cognitive impairment formed a cluster of their own. Furthermore, it seems highly relevant that there should be a profile characterized by cognitive impairment as part of a classification of HA clients. Interventions might have to be structured differently for these individuals due to the impact of cognitive deficits on participation in healthcare planning and self-management [5]. During the process of deciding which cluster-solution best described the sample, the three-cluster solution was ruled out mainly because going from four to three clusters meant that cluster 5 merged with cluster 2. This solution would then have failed to recognize this important group of HA clients.

According to statistics, younger individuals constitute a minority of HA clients in Sweden. Thus, it could be expected that they would classify into smaller clusters of their own: cluster 3 (low level of disability) and cluster 4 (high level of disability). The variable characteristics of cluster 3 were quite close to those of cluster 1 (older adults with low level of disability), except for a considerably higher level of concerns about falling. According to previous research, the prevalence of fear of falling ranges between 21% to as much as 85% among older people living at home [36]. In the sample investigated in this study, 77.8% of the participants declared that they were afraid of falling [37]. A number of disability-related consequences of fear of falling, such as declining physical function and restriction of activity and participation, have been reported among community-dwelling older adults [36]. It is in this population primarily associated with female gender, low physical function, use of a walking aid and a history of falls [38]. The phenomenon seems to be less explored among younger adults, however. It should thus be kept in mind that, despite being quite independent and well-functioning, individuals belonging to this group might have an increased risk of developing disabilities due to concerns about falling.

The participants of cluster 4 were the most dependent and most functionally limited of all the groups. The fall risk associated with transfers and the difficulties they experienced when going to the toilet and taking showers were in line with the group’s high level of concerns about falling as well as their low rating of the usability of their home in the self-care aspect. When comparing the four- and five-cluster solutions, the solutions with five groups was preferred since in the four-cluster solution cluster 4 merged with cluster 2. Even though the majority of HA clients are older it cannot be overlooked that there are also younger persons with a high level of disability in need of HAs [9, 20]. A classification model could consequently not aspire to represent the population of HA clients without including such a group.

The previous classification of this sample of HA clients, including approximately half the number of participants (n = 124), resulted in a six-group solution [21]. When comparing that model with the one presented in the current study there are both similarities and differences. Although alternate concepts were used when labeling the clusters, it is obvious that the structure identified in the original model, with groups differing from each other regarding age, function and disability, is present also in the current one. However, when the number of included participants doubled some changes occurred which need to be discussed.

First, in the six-group model, the clusters were of fairly even size, whereas in the five-group model two of the clusters were considerably larger than the other three. Evidently, as the number of classified participants increased, the groups consisting of older adults grew in size, in relation to the groups with younger adults. The age distribution among the groups in the current model thus seems to be more representative of the population of HA clients in Sweden [9]. Second, age appeared to become less influential in differentiating older adults from each other when the sample size increased. As pointed out earlier, there was in fact hardly any mean age difference at all between the groups comprising of older adults in the five-group model. This suggests that when planning home-based interventions for older adults, the focus should be on variation regarding functional outcomes rather than chronological age. In this context, it is important to note that we did not include any variables related to environmental barriers or characteristics, or the type of HAs conducted in the cluster analysis. One of the aims of identifying a set of client profiles is to achieve guidance in how to evaluate HAs, and to identify those clients in most need of regular monitoring. Also, the profiles may potentially be useful for planning of HAs. Thus, only data related to the person and the activity should be included in the cluster.

A methodological strength of this study is that the cluster analysis was based on data collected with rigorous procedure by experienced and specifically trained professionals [20]. In our study, hierarchical cluster methods were chosen given that their structure is more informative than k-means clustering solutions and they are relatively easy to implement, understand, and reiterate. The typical disadvantage of hierarchical clustering is that decision on the final solution can be arbitrary. This risk was minimized by triangulating the clustering results with the use of qualitative data. A methodological concern that needs to be addressed however is the possible implications of replacing missing data with the sample mean, as explained in the methods section. Common risk when applying missing data replacement with the mean is that some cases might be misclassified due to manual data imputation. However, given the exploratory aim of our study, we evaluated that this risk could be acceptable, given the opportunity to reach a more comprehensive overview of the HA population (and thus, losing to some extent classification accuracy at individual level). Interestingly, the variable with most missing data for this sample was cognitive impairment and plausible reasons for this have been discussed in previous publications [21, 37]. One assumption is that persons with severe cognitive impairment were not able to participate fully in the assessments [21]. This was not unexpected since there certainly are challenges inherent in involving persons with cognitive impairment in this type of research. Unlike the ResHA study, most previous HA studies have not included this sub-group [14], and cognitive impairment of a certain magnitude has been used as exclusion criterion due to anticipated participation difficulties [12, 13]. Replacing the missing MoCA scores of these participants with the sample mean enabled them to be included in the analysis and thus represented in the classification. This could however have affected how they were classified and caused them to end up in cluster 1 (Older adults with low level of disability) or 2 (Older adults with medium/high level of disability) instead of cluster 5 (Older adults with medium level of disability including at least moderate cognitive impairment). Consequently, it should be kept in mind that there might be more participants with severe cognitive impairments in the sample than is reflected by the size of cluster 5.

Our choice of mixed methods design fell on the triangulation design applying a convergence model [24]. The aim was to use the qualitative data to expand and if possible confirm the quantitative results, in order to provide rich examples. A limitation with this choice is that the role of the qualitative approach is more supportive than explorative.

In order to establish credibility of the qualitative results, the analysis process has been thoroughly described and illustrative quotations presented. The purposive sampling of participants with the intention to achieve variation also contributed to strengthen the credibility [33]. Ideally for the validation purpose of this study, participants would not have been sampled for interviews until after the cluster analysis to guarantee an equal number of participants in each group. However, since there was a variation in experience among the interviewed participants, at least one person was classified to each group and the study was possible to conduct with available data.

As for generalizability of the client profiles, the sample does not differ considerably from the population of HA clients in Sweden regarding age and gender distribution. Thus, it is reasonable to conclude that the classification can be applied to other HA clients in Sweden, fulfilling the inclusion criteria for the study. Comparisons with other samples of HA clients can be made based on the presented variable means as well as the demographical information. The five profiles were identified and validated using only cross-sectional data, and whether the cluster solution and the participants’ characteristics are stable or will change over time, remains to be seen from future research applying a longitudinal design.

Conclusions

This classification model of HA clients outlines five client profiles that differ from each other regarding age and level of disability. Applying mixed methods, the validity of the profiles has been strengthened by qualitative data. The identified and validated client profiles should be regarded as another step towards a better understanding of how home-based interventions could be delivered more effectively to groups of HA clients, based on their different characteristics. Knowledge about different characteristics among clients allows for better monitoring over time, according to their specific needs. The study’s mixed methods design might serve as inspiration for future research concerning classification of heterogeneous populations.

Acknowledgements

Sincere thanks are extended to the participants in the study for taking their time, and to the municipality occupational therapists for collecting data. This study was conducted within the context of the Centre for Ageing and Supportive Environments (CASE) at Lund University.

Disclosure statement

No potential conflicts of interest to be reported.

Additional information

Funding

Swedish Research Council for Environment, Agricultural Sciences and Spatial Planning (FORMAS), Sweden Faculty of Medicine at Lund University, Sweden Region Skåne, Sweden Oslo Metropolitan University, OsloMet, Norway.