This paper is in the following e-collection/theme issue:

Email Communication, Web-Based Communication, Secure Messaging (179) Health Services Research (527) Personal Health Records, Patient-Accessible Electronic Health Records, Patient Portals (659) Patient-Clinician Relationship (27) Data Science (319) Precision Medicine (116) Gender Dysphoria and Transgender Issue (21) Natural Language Processing (828)

Published on in Vol 27 (2025)

Preprints (earlier versions) of this paper are available at https://preprints.jmir.org/preprint/64160, first published .
Understanding Gender-Specific Daily Care Preferences: Topic Modeling Study

Understanding Gender-Specific Daily Care Preferences: Topic Modeling Study

Understanding Gender-Specific Daily Care Preferences: Topic Modeling Study

Authors of this article:

Kyungmi Woo1, 2 Author Orcid Image ;   Se Hee Min3 Author Orcid Image ;   Aeri Kim4 Author Orcid Image ;   Subin Choi1 Author Orcid Image ;   Gregory L Alexander5, 6 Author Orcid Image ;   Terrence O’Malley6 Author Orcid Image ;   Maria D Moen7 Author Orcid Image ;   Maxim Topaz5, 8 Author Orcid Image
Article Authors Cited by Tweetations Metrics

    Original Paper

    1College of Nursing, Seoul National University, Seoul, Republic of Korea

    2Visiting Scholar, Columbia University School of Nursing, New York, NY, United States

    3School of Nursing, University of Pennsylvania, Philadelphia, PA, United States

    4The Research Institute of Nursing Science, Seoul National University, Seoul, Republic of Korea

    5School of Nursing, Columbia University, New York, NY, United States

    6Corresponding Faculty, Harvard University, Boston, MA, United States

    7Innovation and External Affairs, MyDirectives Inc, Nashville, TN, United States

    8Columbia University Data Science Institute, New York, NY, United States

    Corresponding Author:

    Kyungmi Woo, PhD

    College of Nursing

    Seoul National University

    103 Daehak-ro, Jongno-gu

    Seoul, 03080

    Republic of Korea

    Phone: 82 2 740 8828

    Fax:82 2 766 1852

    Email: woo2020@snu.ac.kr


    Background: Daily preferences are a reflection of how adults wish to have their needs and values addressed, contributing to joy and satisfaction in their daily lives. Clinical settings often regard older adults as a uniform group, neglecting the diversity within this population, which results in a shortfall of person-centered care that overlooks their distinct daily care preferences. At the heart of person-centered care lies the imperative to comprehend and integrate these preferences into the care process. Recognizing and addressing gender differences in older adults is critical to customizing care plans, thereby optimizing quality of life and well-being for individuals. This study addresses the need to understand the diverse daily care preferences of adults, particularly among older populations, who represent a growing demographic with unique needs and interests.

    Objective: This study aims to identify and analyze the key themes and daily care preferences from unstructured adult text narratives with a focus on uncovering gender-specific variations.

    Methods: This study used 4350 deidentified, unstructured textual data from MyDirectives (MyDirectives, Inc), an interactive online platform. Advanced topic modeling techniques were used to extract meaningful themes, and gender-specific term frequency and distribution were examined to identify gender differences in these elements.

    Results: The study sample included 2883 women (mean age 63.02, SD 13.69 years) and 1467 men (mean age 67.07, SD 11.73 years). Our analysis identified six major themes: (1) “entertainment” (12.14%, 528/4350), (2) “music” (10.39%, 452/4350), (3) “personal interests and memories” (38.18%, 1661/4350), (4) “intimate relationships” (14.92%, 649/4350), (5) “natural comforts” (16.18%, 704/4350), and (6) “emotional, cultural, and spiritual foundations” (8.18%, 356/4350). Gender differences were evident: women were more likely to express preferences for “personal interests and memories” (40.7% vs 33.3%), “natural comforts” (18.4% vs 11.9%), and “emotional and spiritual foundations” (9.3% vs 6.1%) than men. Men expressed stronger preferences for “entertainment” (18.1% vs 9.1%) and “music” (16.8% vs 7.2%). Common terms across all participants included “dog,” “love,” “friends,” and “book.” Notably, the study revealed significant gender differences in daily care preferences, especially regarding familial relationships and entertainment choices.

    Conclusions: The findings underscore the importance of recognizing individual daily care preferences in person-centered care, particularly regarding gender. Understanding these preferences is crucial for improving care quality and patient satisfaction, thereby enhancing the overall quality of life for adults receiving care across our health care system.

    J Med Internet Res 2025;27:e64160

    doi:10.2196/64160

    Keywords

    care preference; patient-centered care; quality of life; data mining; topic modeling; gender differences; online platform; KNIME 


    Older adults live relatively healthier lives with longer life expectancies thanks to recent advances in medical science and technology [1]. However, older adults continue to face significant health care challenges, including a persistent disability burden [2] and reliance on various care services, such as hospitalization and long-term care assistance with daily living activities [3]. Despite these complexities, older adults are often treated as a homogeneous group in clinical settings [4] rather than a heterogeneous group of individuals, leading to a lack of person-centered care that fails to consider their unique daily care preferences [5-7]. This oversight results in suboptimal health outcomes, higher rates of overtreatment and unwanted care [8], and increased patient dissatisfaction [5-7]. Furthermore, the rising number of younger adults in various care settings [9] underscores the importance of understanding daily care preferences across all age groups for delivering high-quality, person-centered care [10].

    Daily preferences reflect how adults prefer their needs and values met for those aspects of living that bring them joy and happiness in their daily lives [11,12]. When daily preferences are fulfilled, adults experience a sense of subjective well-being and happiness, creating motivation to actively engage in life [11,13]. Furthermore, happiness has been associated with a reduced likelihood of all-cause mortality and better health outcomes such as functional independence, especially among older adults [14]. Thus, the tenet of person-centered care is understanding adults’ daily preferences and effectively incorporating them into care [15,16]. Such delivery of person-centered care will enable better health-related quality of life for adults regardless of their care setting [15,16].

    Regardless of age, daily care preferences are highly likely to be individualized and diverse based on individual values, sociodemographic factors, and care settings [11,12]. A recent study found significant gender differences in everyday living preferences among older adults [17]. For example, older women adults had a stronger daily preference in four areas, including having a caregiver of the same gender and a desire to engage in volunteer work, household tasks, and decorating. In contrast, older adults’ men preferred a hands-on approach when learning new things and did not emphasize volunteer work [17]. Another study found a significant difference in future living arrangements for middle-aged adults based on gender [18]. For example, middle-aged women were more likely to prefer to live in a nursing home when they became older, with a lack of preference to live in the community compared with middle-aged men [18]. Recognizing these individual daily care preferences is crucial, and a key aspect to consider is the significant impact of gender on adults’ daily care preferences. Understanding these gender differences among older adults is essential to effectively tailor care plans and ensure the highest quality of life and well-being for adults of all ages [12].

    This retrospective study used deidentified data from MyDirectives, Inc (formerly ADVault, Inc) [19], an interactive and innovative online platform that easily stores, responds to queries, and retrieves health care information between health care providers and patients in an interoperable, standards-based manner. The purpose of MyDirectives, coupled with the mission of the Moving Forward Coalition [20], is to enable the delivery of person-centered care through understanding adults’ daily care preferences and incorporating them into person-centered care plans. The MyDirectives application asks adults to directly describe their daily care preferences in free text, thus providing more abundant and reliable information than simple, categorized responses from the survey data. Using MyDirectives data coupled with topic modeling, a type of natural language processing (NLP) technique, can uncover latent topics within a large corpus of texts that contain information on adults’ daily care preferences [18]. Understanding meaningful topics of daily care preferences can help health care professionals better incorporate them into a plan of care, improving the provision of person-centered care, the quality of patient care, and health outcomes.

    This study identifies and analyzes key themes and daily care preferences from unstructured adult text narratives. Additionally, this analysis explores the variances and commonalities in these themes between genders, offering a comprehensive understanding of daily care preferences as reflected in the narratives.


    Data Source and Sample

    The current study used deidentified, unstructured textual data extracted from MyDirectives, a sizable, interactive online platform facilitating seamless access to health-related information and fostering communication between health care providers and patients. A random sample encompassing 10,000 unique records was derived from community-dwelling adults across the United States. This investigation examined responses to the following targeted query:

    My Likes/Joys: Describe the things that bring you joy. Photographs or other items you would like to have nearby, or music you’d like to hear. A favorite pillow, a night light, or your favorite flowers.

    Furthermore, participant gender and age (in 10-year increments) were acquired by administering standardized questionnaires embedded within the MyDirectives interface.

    Data Preparation

    From 10,000 records, 4646 records were processed after excluding 5331 records with missing responses in the daily care preferences, 1 record with missing gender or birth year information, and 22 records with gender answered as “transgender” (N=9) or “not specified” (N=13). The exclusion of records from the transgender group (N=9) was due to the small sample size, which would not provide sufficient data for robust analysis using topic modeling techniques and gender differences. In addition, 296 records with irrelevant responses (eg, “N/A” and “None”) were removed. As a result, a final dataset of 4350 records was used for the latent Dirichlet allocation (LDA) topic modeling (Figure 1).

    ‎
    Figure 1. Flowchart illustrating the data selection process for the topic modeling study on daily care preferences among US adults. The figure outlines participant inclusion criteria, exclusion steps, and the final sample (N=4350) used for natural language processing analysis.

    Preprocessing

    In this study, we applied standard NLP procedures to prepare textual data for analysis. The process involved several steps:

    1. Removal of punctuation and numerical characters: We stripped away all punctuation marks and numbers from the text, as these elements are not required for our NLP analysis.
    2. Conversion to lowercase: The entire text was converted to lowercase to ensure uniformity, so the system treats words such as “Health” and “health” the same.
    3. Tokenization: This step involved breaking down the text into individual elements, or “tokens,” typically words or phrases.
    4. Lemmatization: We transformed words into their base or root form. For instance, variations of a word such as “running” were reduced to their simplest form, “run.”
    5. Elimination of common English stop words: Commonly used English words such as “I,” “my,” “on,” “of,” “what,” “and,” “the,” and “from” were removed. These words, often called stop words, are usually excluded in NLP as they offer little value in understanding the key topics of the text.

    Topic Modeling

    The study used LDA topic modeling using the Konstanz Information Miner (KNIME) Analytics Platform (version 4.7.2; KNIME) [21]. Python (version 3.10.12; Python Software Foundation) was used to determine the most appropriate number of topics. LDA topic modeling, an unsupervised machine learning technique and a generative probabilistic algorithm, aims to uncover the number of topics (K) within extensive document collections [22]. A random mixture of latent topics represents these documents, defining the distribution of topics within a specific document (α) and the distribution of words for a given topic (β) [22,23]. The latent topic themes are not directly observable and thus necessitate inference based on their constituent words. Researchers specify K as input hyperparameters (Multimedia Appendix 1). This study determined the optimal K, as described in the “Choice of Optimal K for LDA Topic Modeling” section [24,25]. Previous studies have effectively used similar parameters in LDA topic modeling to reveal concealed themes across diverse research domains within large corpora, including health care–related patient feedback, newspapers, and open-ended survey responses [26-28]. An expert panel of 7 specialists in relevant fields comprehensively assessed the topic modeling results. The panel included 4 researchers (KW, AK, MT, and SC) with expertise in topic modeling, 3 (KW, SHM, and AK) with a background in psychological health research among older adults, and 1 (KW) with experience in care planning. In addition, we had experts (KW and AK) with a background in nursing home care, MT specializing in informatics, and GLA, TAO, and MDM, who are authorities in nursing homes and long-term care and are leading efforts in nursing home reform through the Moving Forward Coalition.

    Choice of Optimal K for LDA Topic Modeling

    To ensure our study was accurate, we needed to determine the best number of topics to use in our analysis, a process known in technical terms as selecting the optimal “K” for LDA topic modeling. This step is crucial because too many topics can make the analysis confusing and overlapping, while too few can make the topics too broad and unclear. We used several methods to find the right balance:

    1. Perplexity measure: This helped us understand how well our model could predict the text. We looked for the lowest perplexity, meaning the model was more accurate in understanding the text [29].
    2. Coherence score: Using Python Gensim (RARE Technologies Ltd) [30], we measured how logically the words within each topic related. Higher scores meant the topics made more sense, such as the words “nurse,” “care,” “patient,” and “hospital” being grouped.
    3. Elbow method: We also used the elbow method, which helped us visualize the best number of topics by looking at how texts grouped at different numbers of topics [31].

    By combining these methods, we could find the most effective number of topics for our study, ensuring our analysis was accurate and understandable [32].

    Gender Differences Throughout the Keywords and Themes

    A term frequency analysis was used to identify gender differences in the usage and prevalence of keywords associated with daily care preferences. We extracted and normalized the frequencies of the top 40 words for each gender, and these normalized values were visualized in heatmaps, allowing for a visual comparison of term usage priorities and intensities between genders. Notably, the prominence of the top 4 words (“music,” “family,” “photo,” and “picture”) in both men’s and women’s datasets was primarily due to their inclusion as examples within the My Likes/Joys open-ended question. Consequently, their consistently high occurrence dampened the variability in the heatmap. Therefore, after excluding these 4 terms, we investigated gender differences in daily care preferences by comparing the heatmaps generated from the remaining top 40 words. Figure 2 represents the procedural workflow executed in KNIME, outlining the steps of the actual research. In addition, a radar chart, also called a spider chart, is used to visualize the comparison of gender distributions across themes, providing a comprehensive view of gender patterns and differences on a single chart. A radar chart displays each value of a variable along axes arranged at equal angles around a central point. Within this structure, each variable is plotted along its corresponding axis, forming a polygon, enabling the comparison and visualization of multivariate data. The central point represents the smallest value and extends outward to represent larger ones.

    ‎
    Figure 2. The Konstanz Information Miner workflow of topic modeling applied to unstructured text responses collected via MyDirectives. The process includes tokenization, lemmatization, and removal of stopwords before topic modeling using latent Dirichlet allocation. Tokenization refers to the process of breaking down a stream of text into smaller units called tokens, typically words. Lemmatization is the process of converting variations of a word that reflect tense or number to its root form, called a lemma. For instance, the words “writes,” “wrote,” and “written” would all be lemmatized to “write.” Stopwords are words that are deemed to have little to no intrinsic value for understanding the essence of the content. These commonly include words such as “I,” “my,” “on,” “of,” “what,” “and,” “the,” and “from.” Model optimization was guided by perplexity, coherence, and the elbow method. LDA: latent Dirichlet allocation.

    Ethical Considerations

    This study was reviewed and deemed exempt from institutional review board review by Columbia University (exemption number AAAU5320), under ethical guidelines for secondary analysis of deidentified data. As the dataset consisted of deidentified data provided under a formal data use agreement, no additional informed consent was required for this analysis. The original data collection obtained appropriate consent from the users, and the secondary use was authorized by the data provider, MyDirectives, Inc. To protect participant privacy and confidentiality, all data were fully deidentified before analysis, and no personally identifiable information was available to the research team at any stage. No compensation was provided, as this study did not involve direct contact with participants. Finally, no images or figures in the paper or Multimedia Appendix 1 contain any information that could be used to identify individual participants.


    Participant Characteristics and Text Data

    The LDA topic modeling analyzed 4350 free-text responses regarding daily care preferences among adults. The average age of the 4350 participants was 64.38 (SD 13.20) years, consisting of 2883 women (mean age 63.02, SD 13.69 years) and 1467 men (mean age 67.07, SD 11.73 years). Regarding the free-text data, the average word count was 18.92 (SD 24.11) with an average character count of 107.18 (SD 131.90).

    Optimal Number of Topics

    The determination of the optimal number of topics involved several evaluation methods, including perplexity, coherence, and the elbow method. Figure 3 shows the results of these evaluation approaches. Although K=4 shows the lowest perplexity score, it also yields the lowest coherence score (Figure 3A), suggesting potential inconsistencies and reduced interpretability within the model. As K increased to K=8, the topic coverage declined while consistency improved. The elbow method illustrated a sharp shift in the sum of squared errors values as K transitioned from 3 to 4, remaining stable as K ranged between 5 and 7 (Figure 3B). Considering these findings, individual LDA topic models were constructed at K=5, 6, and 7, each comprised of 15 keywords. The research team conducted iterative reviews to evaluate interpretability and distinguishability, concluding that 6 topics were the most appropriate. For the LDA topic model with 6 topics, α (0.31) and β (0.91) values were selected as hyperparameters, exhibiting the highest coherence.

    ‎
    Figure 3. Determining the optimal number of topics for latent Dirichlet allocation modeling. (A) Coherence and perplexity scores across topic counts. (B) Elbow method illustrating model fit. Analyses were conducted using Python and Konstanz Information Miner.

    Topic Modeling

    Table 1 displays the identified themes, their top 15 keywords, and the respective proportions for each topic, extracted through LDA topic modeling of the free-text responses regarding daily care preferences in care among adults. The multidisciplinary team collaboratively identified six dominant themes: (1) “entertainment,” (2) “music,” (3) “personal interests and memories,” (4) “close relationships,” (5) “natural comforts,” and (6) “emotional and spiritual grounding.” Notably, the results showed that the theme of “Personal interests and memories” (Topic 3) accounted for the largest proportion of daily care preferences at 38.18%. “natural comforts” (topic 5) and “close relationships” (Topic 4) followed with 16.18% and 14.92%, respectively.

    Table 1. Latent Dirichlet allocation topic modeling results based on 4350 free-text responses from US adults using MyDirectives, an online advance care planning platform.
    RankTopic 1: EntertainmentTopic 2: MusicTopic 3: Personal interests and memoriesTopic 4: Close relationshipsTopic 5: Natural comfortsTopic 6: Emotional and spiritual grounding

    		KeywordWeight (528/4350, 12.14%)KeywordWeight (452/4350, 10.39%)KeywordWeight (1661/4350, 38.18%)KeywordWeight (649/4350, 14.92%)KeywordWeight (704/4350, 16.18%)KeywordWeight (356/4350, 8.18%)
    1Music500Music1063Music2915Family568Music712Music1740
    2Book429Country456Family2502Picture508Flower418Flower367
    3Movie379Rock221Picture1145Photo446Dog350Love182
    4Cat180Jazz151Photo1020Dog420Pillow325Dog123
    5Phone144Song102Flower818Friend359Time312Time122
    6Reading109Piano91Dog374Child353Blanket247Gospel116
    7Video105Roll81Pillow371Time340Animal236Joy109
    8Game95Blues74Friend349Grandchild300Window228Rose100
    9Sport92Pop67Country347Joy239Bed198Song98
    10Computer87Pink65Gospel332Husband208Nature193Life92
    11Radio86Guitar54Pet281Son195Sound152People80
    12Football86John49Photograph273Daughter192Light148Day78
    13Comedy81Play49Rose242Wife172Bird147Color75
    14Puzzle75Folk46Kid232Life172Ocean138God54
    15Access70Beethoven46Night light220Sister162Air131Food51

    The first theme was identified as “entertainment.” The keywords, such as “music,” “book,” “movie,” “video,” “game,” “sport,” “football,” “comedy,” and “puzzle,” indicate the diverse range of entertainment preferences. This theme involves various activities, including listening to music, reading books, watching movies, and solving puzzles. Additionally, the inclusion of keywords such as “phone,” “computer,” and “radio” indicates the technological media used to access entertainment.

    The analysis revealed the second theme, labeled “music.” Notably, “music” stands out as the most prominent keyword across all themes, except “close relationships” (Topic 4). The extracted keywords represent various music genres, including “country,” “rock,” “jazz,” “rock and roll,” “blues,” and “pop.” Specific musical instruments such as “piano” and “guitar” are also mentioned alongside favored composers and singers. Hence, this theme highlights a broad spectrum of musical interests, genres, instruments, and preferences for artists or composers.

    The third theme, “personal interests and memories,” accounts for the largest proportion of all topics. It includes a wide range of personal interests, such as “music,” “flowers,” “dogs,” “pillows,” and “night lights.” Moreover, it emphasizes the importance of valuing memories, indicated by keywords related to “family,” “friends,” “pets,” and “kid,” revealing a wish to preserve these moments through “photographs.”

    The fourth theme, “close relationships,” explores the importance of emotional ties with loved ones. It includes keywords such as “family,” “dog,” “friend,” “child,” “grandchild,” “husband,” “son,” “daughter,” “wife,” and “sister.” This theme encapsulates the essence of cherished relationships, underscoring a preference to share moments of “joy” and precious “time” with close individuals.

    The fifth theme, “natural comforts,” focuses on seeking solace, tranquility, and comfort within nature. This theme is characterized by nature-related keywords, including “flower,” “dog,” “animal,” “light,” “sound,” “bird,” “ocean,” and “air,” which convey a sense of the natural world and emphasize a preference for the tranquil atmosphere they provide. Additionally, it indicates a preference for comfort when keywords such as “music,” “pillows,” “light,” and “blankets” are accompanied by “time,” “bed,” and “window.”

    Finally, the sixth theme, “emotional and spiritual grounding,” is centered around keywords such as “music,” “love,” “dog,” “time,” “gospel,” “joy,” “song,” “people,” “day,” and “god.” This theme reveals the significance of emotional connections by emphasizing love and joy, portrayed through music, companionship with pets, and interactions with people. Furthermore, it extends into the spiritual realm, bringing out the essence of joy and love through the gospel and God.

    Gender Differences in Daily Care Preferences

    The gender differences in daily care preferences associated with a better quality of life were examined and displayed in Figure 4A, using the frequency of the top 40 words by each gender in the sample. The top four words, “music,” “family,” “photo,” and “picture,” were predominant in both genders. Except for the top four words, “dog,” “love,” “friends,” and “book,” they were the common, frequent top 10 words in men and women. Men more commonly mentioned hobby-related words such as “movie,” “sports,” “fishing,” “computer,” and “television,” while for women, “blanket,” “joy,” and “phone” were more prevalent. These findings were further supported by the results shown in the radar chart in Figure 4B. Women demonstrated higher daily care preferences for themes associated with “personal interests and memories,” “natural comforts,” and “emotional and spiritual grounding” than men. meanwhile, men preferred “entertainment” and “music” themes more than women.

    ‎
    Figure 4. Gender-based keywords and thematic differences in daily care preferences among US adults. (A) Heat map of normalized term frequencies for the top 40 most common words by gender. (B) Radar chart comparing the distribution of 6 major themes across women and men subgroups, based on topic probability weights.

    Principal Findings

    To the best of our knowledge, this is the first study to identify keywords and themes that capture the daily care preferences of adults residing in the United States, specifically focusing on examining gender differences based on word frequency and themes. Using LDA topic modeling, we extracted six distinct themes: (1) “entertainment,” (2) “music,” (3) “personal interests and memories,” (4) “close relationships,” (5) “natural comforts,” and (6) “emotional and spiritual grounding.” in addition, we found gender differences in specific familial relationships (eg, “close relationships”) and their preferred entertainment activities.

    The first theme, “entertainment,” emerged as a broad topic in our study, incorporating traditional activities such as reading, sports, and music. About gender differences within this theme, it was notable that the entertainment theme accounted for nearly twice the proportion of daily care preferences for men compared with women (18.13% vs 9.09%). This marked the second-largest gender difference among the themes. Moreover, solitary activities such as watching movies and reading books were more prominent in men than women. In contrast, women exhibited a stronger tendency toward interpersonal connections, expressing emotions (ie, love and joy) through words. The “Entertainment” theme, which encompasses technological terms such as “phone,” “video,” “game,” and “computer,” suggests a significant digital dimension to entertainment activities. Recent studies exploring technology in adult entertainment, including virtual reality, have suggested potential positive impacts on physical and psychological health [33,34]. The current study noted a clear preference for electronic devices such as computers, televisions, and phones among men compared with women. Meanwhile, in the context of nursing homes, research has shown that men residents were less satisfied with the activities provided to them than women, indicating a significant disparity in satisfaction between genders [35]. This disparity may suggest a lack of integration of gender-specific entertainment preferences into care and care plans. Consequently, it is important to consider gender differences when providing entertainment in adult care, as supported by studies demonstrating varied attitudes toward communication and entertainment devices between genders [36,37].

    The music theme encompassed various genres, indicating a broad spectrum of musical preferences among adults. This thematic prominence of “music” within the adult preference spectrum has significant implications for person-centered care, especially in nursing homes. Recognizing the diversity of musical interests can guide the development of personalized care plans that use music therapeutically to help improve health status by reducing depression, agitation, behavior problems, blood pressure, anxiety, and pain [38-40]. The perceived importance of music in the relationship between music and pain management could serve as a mechanism for better pain management, care planning, and quality of life [40]. Gendered preferences within this music theme were particularly pronounced: men showed a stronger tendency toward “rock music,” while women showed a preference for “country songs.” These results revealed that “music” had a significantly higher share of interest among men than women in their preferences, and the distinct music styles preferred by men and women align with findings from existing research [40,41]. Based on this finding, care providers may need to assess and provide tailored musical activities to enhance health-related quality of life for nursing home residents.

    The third theme, “personal interests and memories,” constitutes the majority of responses, underscoring its significance for both men and women as the most meaningful aspect of their daily care preferences. Accordingly, it becomes evident that integrating individual preferences and histories into daily care is essential, ranging from music and family to photos, botanical elements, beloved animals, and comforting items such as pillows and night lights. More specifically, in a person-centered care framework, Morgan and Yoder [42] mentioned that recognizing each resident’s unique histories and preferences and incorporating these individualized elements into care plans are crucial. Recognizing preferences to support person-centered care may help to improve a sense of dignity, autonomy, and quality of care while reducing behavioral disturbances [42-44]. When individuals encounter familiar objects and themes from their personal histories, such as a favorite family photo or a beloved flower, it can stimulate memories and emotions, enhancing cognitive engagement and emotional satisfaction [45]. Notably, our findings highlight a gendered pattern in these interests and memories, where men were prominently associated with keywords such as music, dogs, and photographs, while women were more connected to keywords suggesting a familiar atmosphere, such as flowers, pillows, blankets, and night lights. By considering these preferences, staff can cultivate a home-like care environment that not only meets physical health needs but also nourishes the psychological, emotional, and unique aspects of a person’s life [46].

    The fourth topic, “close relationships,” highlights the essence of human connection through keywords such as family, friends, and pets. Interestingly, while “family” emerged as an important element across genders, there were notable differences in the priority of specific family members. Men frequently mentioned “wife,” whereas women placed “child” and “grandchild” above “husband” in prominence. This suggests gender-based differences in the perception and valuation of family roles and relationships [47,48]. In long-term care contexts, these insights are vital for person-centered care, as families often play an important role in care planning. Care plans that involve family members not only reduce depression but also enhance their sense of well-being and belonging [46,49]. For men, activities that facilitate communication with their spouse or celebrate anniversaries may be beneficial, whereas for women, opportunities to interact with children or grandchildren or share family stories may be particularly impactful.

    The fifth topic, “natural comforts,” characterized by keywords such as music, flowers, animals, and natural elements, showed a profound preference for peace and relaxation, emphasizing the importance of surrounding oneself with nature’s tranquility and comfortable environments. This theme held varying degrees of importance among gender-specific clusters, ranking fifth in the men’s cohort and second in the women’s cohort within the themes identified. These findings underscore the need for gender-sensitive care planning of adult daily care preferences, particularly in their care environments. For women, this theme indicated a profound value placed on tranquility and a connection to nature. Integrating features such as garden-related activities and providing rooms with natural views could improve their quality of life and psychological health [50,51]. In contrast, men might prefer different expressions of natural comforts, such as pet therapy or the sounds of a river, which offer a calming and mood-enhancing effect for those who enjoy fishing [52]. Moreover, care strategies could include access to outdoor spaces or nature-themed outdoor activities that align with these identified daily care preferences.

    The sixth theme, “emotional and spiritual grounding,” which recognizes humans as spiritual beings, is a critical aspect, despite its relatively low representation in the data. Previous studies have emphasized the importance of emotional and spiritual support for adults, as evidenced by well-being, religion, and comfort [53-55]. A recent mixed methods study on older nursing home residents highlighted the significance of religion and emotional support in their lives [56], further emphasizing the enduring importance of these themes across different settings. Therefore, it is important to understand and integrate an individual’s religious and cultural beliefs and values into the person-centered care plan and care delivery. Introducing care environment elements and tailored activities that align with this theme, such as organizing music sessions with spiritually uplifting songs, establishing areas for reflection or prayer, offering life review programs, or arranging pastoral visits, can contribute to improving adults’ life satisfaction, spiritual well-being, connectedness, and discovery of meaning in life [57,58].

    Limitations

    While our exploratory study offered meaningful insights, several limitations should be considered. First, efforts to stratify the sample by gender were hindered by the shortness and small size of the men’s textual data, resulting in pronounced sparsity and challenges in conducting reliable gender-based topic modeling. Consequently, we relied on word frequency analysis to detect patterns and the most discussed topics or concepts within each gender. Furthermore, despite recognizing the importance of including the perspective of transgender individuals in understanding daily care preferences, the limited number of responses from this group did not meet the substantial data requirements necessary for comprehensive topic modeling and examination of gender differences. Future investigations should encompass a larger and more diverse dataset to facilitate a more detailed and comprehensive analysis across the gender spectrum. While our paper mostly represents older adults, expanding the scope to a larger dataset offers opportunities to explore the daily care preferences of young and middle-aged adults, potentially yielding valuable insights through comparative studies. It is crucial to acknowledge that our findings are shaped by the available data and sample size and thus may not be generalizable to different age groups. Recent research endeavors are pivotal to refining our understanding and laying the groundwork for developing personalized care plans that account for individual daily care preferences. In addition, this research identified 6 themes related to the daily care preferences of adults, explored gender differences through the proportion of revealed themes from topic modeling and frequently occurring words in the text data, and recommended building person-centered care plans considering gender differences. However, for more refined person-centered care, future research could collect additional data features, such as health status, functional status, communication preferences, and personal history, to build and use machine learning models to create customized care plans.

    Conclusions

    This study explored the daily care preferences of community-dwelling adults and uncovered specific environmental factors that play a crucial role. The identified themes of daily care preferences empower clinicians to enhance health-related quality of life for individuals, allowing for incorporating these insights into personalized care planning. Moreover, recognizing gender differences is also essential when developing person-centered care plans. With these findings, the feasibility of care planning based on personal daily care preferences may be enhanced, further emphasizing the ongoing need to tailor care for adults in the community.

    Acknowledgments

    AK and SC received a scholarship from the BK21 education program (Center for Human-Caring Nurse Leaders for the Future).

    Data Availability

    The datasets analyzed during this study are not publicly available due to confidentiality and proprietary restrictions, but are available from the corresponding author on reasonable request and with permission from MyDirectives, Inc.

    Authors' Contributions

    KW supervised and administered the project. KW, SHM, and MT designed and conceptualized the study. GLA played a pivotal role in facilitating collaboration with MyDirectives and the Moving Forward Coalition. GLA, TAO, and MDM acquired and shared the deidentified MyDirectives data. KW and AK produced the core software code. KW, SHM, AK, and SC analyzed and interpreted the results of the data. KW, SHM, and AK wrote the original manuscript. GLA, TAO, MDM, and MT contributed to manuscript editing and provided critical feedback. All authors reviewed and approved the final version of the manuscript.

    Conflicts of Interest

    None declared.

    Multimedia Appendix 1

    The graphical model for latent Dirichlet allocation. D-plate = For all D document, k-plate = For all K topics, N-plate = For all N words, α = Hyperparameter of the Dirichlet distribution, θ_d = Per-document topic proportion, Z_(d,n) = Per-word topic assignment in document d, w_d = Observed word for document d, β_k = Distributions of the words for a given topic K, η= Hyperparameter of the Dirichlet distribution prior on the β_k.

    DOCX File , 98 KB
    1. Crimmins EM. Lifespan and healthspan: past, present, and promise. Gerontologist. 2015;55(6):901-911. [FREE Full text] [CrossRef] [Medline]
    2. GBD 2019 Ageing Collaborators. Global, regional, and national burden of diseases and injuries for adults 70 years and older: systematic analysis for the global burden of disease 2019 study. BMJ. 2022;376:e068208. [FREE Full text] [CrossRef] [Medline]
    3. de Jong L, Zeidler J, Damm K. A systematic review to identify the use of stated preference research in the field of older adult care. Eur J Ageing. 2022;19(4):1005-1056. [FREE Full text] [CrossRef] [Medline]
    4. Tobis S, Jaracz K, Kropińska S, Talarska D, Hoe J, Wieczorowska-Tobis K, et al. Needs of older persons living in long-term care institutions: on the usefulness of cluster approach. BMC Geriatr. 2021;21(1):316. [FREE Full text] [CrossRef] [Medline]
    5. Kogan AC, Wilber K, Mosqueda L. Person-centered care for older adults with chronic conditions and functional impairment: a systematic literature review. J Am Geriatr Soc. 2016;64(1):e1-e7. [CrossRef] [Medline]
    6. Tiffen J, Corbridge SJ, Slimmer L. Enhancing clinical decision making: development of a contiguous definition and conceptual framework. J Prof Nurs. 2014;30(5):399-405. [CrossRef] [Medline]
    7. Tinetti ME, Costello DM, Naik AD, Davenport C, Hernandez-Bigos K, Van Liew JR, et al. Outcome goals and health care preferences of older adults with multiple chronic conditions. JAMA Netw Open. 2021;4(3):e211271. [FREE Full text] [CrossRef] [Medline]
    8. Tackling overtreatment and overspending in U.S. health care. The Commonwealth Fund. 2023. URL: https:/​/www.​commonwealthfund.org/​publications/​podcast/​2023/​nov/​tackling-overtreatment-overspending-us-health-care [accessed 2024-02-12]
    9. Ne'eman A, Stein M, Grabowski DC. Nursing home residents younger than age sixty-five are unique and would benefit from targeted policy making. Health Aff (Millwood). 2022;41(10):1449-1459. [CrossRef] [Medline]
    10. Laws MB, Beeman A, Haigh S, Wilson IB, Shield RR. Prevalence of serious mental illness and under 65 population in nursing homes continues to grow. J Am Med Dir Assoc. 2022;23(7):1262-1263. [FREE Full text] [CrossRef] [Medline]
    11. Van Haitsma K, Abbott K, Arbogast A, Bangerter L, Heid A, Behrens L, et al. A preference-based model of care: an integrative theoretical model of the role of preferences in person-centered care. Gerontologist. 2020;60(3):376-384. [FREE Full text] [CrossRef] [Medline]
    12. Van Haitsma K, Abbott KM, Heid AR, Carpenter B, Curyto K, Kleban M, et al. The consistency of self-reported preferences for everyday living: implications for person-centered care delivery. J Gerontol Nurs. 2014;40(10):34-46. [FREE Full text] [CrossRef] [Medline]
    13. CHEN T. Living arrangement preferences and realities for elderly Chinese: implications for subjective wellbeing. Ageing Soc. 2018;39(8):1557-1581. [CrossRef]
    14. Chei CL, Lee JM, Ma S, Malhotra R. Happy older people live longer. Age Ageing. 2018;47(6):860-866. [CrossRef] [Medline]
    15. Hurtado MP, Swift EK, Corrigan JM. Envisioning the National Health Care Quality Report. Washington, DC. National Academies Press; 2001.
    16. Sepucha K, Ozanne EM. How to define and measure concordance between patients' preferences and medical treatments: a systematic review of approaches and recommendations for standardization. Patient Educ Couns. 2010;78(1):12-23. [CrossRef] [Medline]
    17. Van Haitsma K, Curyto K, Spector A, Towsley G, Kleban M, Carpenter B, et al. The preferences for everyday living inventory: scale development and description of psychosocial preferences responses in community-dwelling elders. Gerontologist. 2013;53(4):582-595. [CrossRef] [Medline]
    18. Qin M, Falkingham J, Evandrou M, Vlachantoni A. Attitudes and preferences towards future old-age support amongst tomorrow’s elders in China. DemRes. 2020;43:285-314. [CrossRef]
    19. MyDirectives. URL: https://www.mydirectives.com [accessed 2023-09-07]
    20. Addressing Residents' Goals, Preferences and Priorities. Moving Forward. 2023. URL: https:/​/movingforwardcoalition.​org/​wp-content/​uploads/​2023/​07/​Addressing-Residents-Goals-Preferences-and-Priorities.​pdf [accessed 2023-09-10]
    21. Ihrmark D, Tyrkkö J. Learning text analytics without coding? An introduction to KNIME. EFI. 2023;39(2):121-137. [CrossRef]
    22. Blei DM. Probabilistic topic models. Commun. ACM. 2012;55(4):77-84. [CrossRef]
    23. Chauhan U, Shah A. Topic modeling using latent dirichlet allocation: a survey. ACM Comput Surv. 2021;54(7):1-35. [CrossRef]
    24. Wang L, Lakin J, Riley C, Korach Z, Frain L, Zhou L. Disease trajectories and end-of-life care for dementias: latent topic modeling and trend analysis using clinical notes. AMIA Annu Symp Proc. 2018;2018:1056-1065. [FREE Full text] [Medline]
    25. Blei D, Ng A, Jordan M. Latent dirichlet allocation. J Mach Learn Res. 2003;3:993-1022. [FREE Full text] [CrossRef]
    26. Liu Q, Chen Q, Shen J, Wu H, Sun Y, Ming WK. Data analysis and visualization of newspaper articles on thirdhand smoke: a topic modeling approach. JMIR Med Inform. 2019;7(1):e12414. [FREE Full text] [CrossRef] [Medline]
    27. Fairie P, Zhang Z, D'Souza AG, Walsh T, Quan H, Santana MJ. Categorising patient concerns using natural language processing techniques. BMJ Health Care Inform. 2021;28(1):e100274. [FREE Full text] [CrossRef] [Medline]
    28. Buenano-Fernandez D, Gonzalez M, Gil D, Lujan-Mora S. Text mining of open-ended questions in self-assessment of university teachers: an LDA topic modeling approach. IEEE Access. 2020;8:35318-35330. [CrossRef]
    29. Churchill R, Singh L. The evolution of topic modeling. ACM Comput Surv. 2022;54(10s):1-35. [CrossRef]
    30. Röder M, Both A, Hinneburg A. Exploring the space of topic coherence measures. 2015. Presented at: WSDM '15: Proceedings of the Eighth ACM International Conference on Web Search and Data Mining; February 2-6, 2015:399-408; Shanghai, China. [CrossRef]
    31. Steinskog AO, Therkelsen JF, Gambäck B. Twitter topic modeling by tweet aggregation. 2017. Presented at: Proceedings of the 21st Nordic Conference on Computational Linguistics; May 22-24, 2017:77-86; Gothenburg, Sweden. URL: https://aclanthology.org/W17-0210.pdf
    32. Kim A, Choi S, Woo K. Navigating the landscape of telemedicine research: a topic modeling approach for the present and future. Telemed J E Health. 2024;30(5):1378-1393. [CrossRef] [Medline]
    33. Peng X, Menhas R, Dai J, Younas M. The COVID-19 pandemic and overall wellbeing: mediating role of virtual reality fitness for physical-psychological health and physical activity. Psychol Res Behav Manag. 2022;15:1741-1756. [FREE Full text] [CrossRef] [Medline]
    34. Lai X, Lei X, Chen X, Rau PLP. Can virtual reality satisfy entertainment needs of the elderly? The application of a VR headset in elderly care. 2019. Presented at: Proceeding of the 21st International Conference on Human-Computer Interaction; July 26-31, 2019:159-172; Orlando, FL. [CrossRef]
    35. Davila H, Ng W, Akosionu O, Thao MS, Skarphol T, Virnig BA, et al. Why men fare worse: a mixed-methods study examining gender differences in nursing home resident quality of life. Gerontologist. 2022;62(9):1347-1358. [FREE Full text] [CrossRef] [Medline]
    36. Halmdienst N, Radhuber M, Winter-Ebmer R. Attitudes of elderly Austrians towards new technologies: communication and entertainment versus health and support use. Eur J Ageing. 2019;16(4):513-523. [FREE Full text] [CrossRef] [Medline]
    37. Van Volkom M, Stapley JC, Malter J. Use and perception of technology: sex and generational differences in a community sample. Educ Gerontol. 2013;39(10):729-740. [CrossRef]
    38. Lorber M, Divjak S. Music therapy as an intervention to reduce blood pressure and anxiety levels in older adults with hypertension: a randomized controlled trial. Res Gerontol Nurs. 2022;15(2):85-92. [CrossRef] [Medline]
    39. Mileski M, Brooks M, Kirsch A, Lee F, LeVieux A, Ruiz A. Positive physical and mental outcomes for residents in nursing facilities using music: a systematic review. Clin Interv Aging. 2019;14:301-319. [FREE Full text] [CrossRef] [Medline]
    40. Mitchell LA, MacDonald RA, Knussen C, Serpell MG. A survey investigation of the effects of music listening on chronic pain. Psychol Music. 2007;35(1):37-57. [CrossRef]
    41. Rentfrow PJ, Goldberg LR, Levitin DJ. The structure of musical preferences: a five-factor model. J Pers Soc Psychol. 2011;100(6):1139-1157. [FREE Full text] [CrossRef] [Medline]
    42. Morgan S, Yoder LH. A concept analysis of person-centered care. J Holist Nurs. 2012;30(1):6-15. [CrossRef] [Medline]
    43. Li J, Porock D. Resident outcomes of person-centered care in long-term care: a narrative review of interventional research. Int J Nurs Stud. 2014;51(10):1395-1415. [CrossRef] [Medline]
    44. Paddock K, Brown Wilson C, Walshe C, Todd C. Care home life and identity: a qualitative case study. Gerontologist. 2019;59(4):655-664. [FREE Full text] [CrossRef] [Medline]
    45. Villar F, Serrat R. Changing the culture of long-term care through narrative care: individual, interpersonal, and institutional dimensions. J Aging Stud. 2017;40:44-48. [CrossRef] [Medline]
    46. Davies M, Zúñiga F, Verbeek H, Simon M, Staudacher S. Exploring interrelations between person-centered care and quality of life following a transition into long-term residential care: a meta-ethnography. Gerontologist. 2023;63(4):660-673. [FREE Full text] [CrossRef] [Medline]
    47. Akiyama H, Elliott K, Antonucci TC. Same-sex and cross-sex relationships. J Gerontol B Psychol Sci Soc Sci. 1996;51(6):P374-P382. [CrossRef] [Medline]
    48. Gurung RAR, Taylor SE, Seeman TE. Accounting for changes in social support among married older adults: insights from the MacArthur Studies of Successful Aging. Psychol Aging. 2003;18(3):487-496. [CrossRef] [Medline]
    49. Tan JDL, Maneze D, Montayre J, Ramjan LM, Wang D, Salamonson Y. Family visits and depression among residential aged care residents: an integrative review. Int J Nurs Stud. 2023;146:104568. [FREE Full text] [CrossRef] [Medline]
    50. van der Velde-van Buuringen M, Hendriks-van der Sar R, Verbeek H, Achterberg WP, Caljouw MAA. The effect of garden use on quality of life and behavioral and psychological symptoms of dementia in people living with dementia in nursing homes: a systematic review. Front Psychiatry. 2023;14:1044271. [FREE Full text] [CrossRef] [Medline]
    51. Schweitzer M, Gilpin L, Frampton S. Healing spaces: elements of environmental design that make an impact on health. J Altern Complement Med. 2004;10 Suppl 1:S71-S83. [CrossRef] [Medline]
    52. Mueller MK, King EK, Callina K, Dowling-Guyer S, McCobb E. Demographic and contextual factors as moderators of the relationship between pet ownership and health. Health Psychol Behav Med. 2021;9(1):701-723. [FREE Full text] [CrossRef] [Medline]
    53. Connolly M, Timmins F. Spiritual care for individuals with cancer: the importance of life review as a tool for promoting spiritual well-being. Semin Oncol Nurs. 2021;37(5):151209. [FREE Full text] [CrossRef] [Medline]
    54. Agli O, Bailly N, Ferrand C. Spirituality and religion in older adults with dementia: a systematic review. Int Psychogeriatr. 2015;27(5):715-725. [FREE Full text] [CrossRef] [Medline]
    55. Naik AD, Martin LA, Moye J, Karel MJ. Health values and treatment goals of older, multimorbid adults facing life-threatening illness. J Am Geriatr Soc. 2016;64(3):625-631. [FREE Full text] [CrossRef] [Medline]
    56. Ludlow K, Churruca K, Mumford V, Ellis LA, Braithwaite J. Aged care residents' prioritization of care: a mixed-methods study. Heal Expect. 2021;24(2):525-536. [FREE Full text] [CrossRef] [Medline]
    57. Dos Santos FC, Macieira TGR, Yao Y, Hunter S, Madandola OO, Cho H, et al. Spiritual interventions delivered by nurses to address patients' needs in hospitals or long-term care facilities: a systematic review. J Palliat Med. 2022;25(4):662-677. [FREE Full text] [CrossRef] [Medline]
    58. Ødbehr LS, Hauge S, Danbolt LJ, Kvigne K. Residents' and caregivers' views on spiritual care and their understanding of spiritual needs in persons with dementia: a meta-synthesis. Dementia (London). 2017;16(7):911-929. [CrossRef] [Medline]

    ‎
    K: number of topics
    KNIME: Konstanz Information Miner
    LDA: latent Dirichlet allocation
    NLP: natural language processing

    Edited by A Mavragani; submitted 10.07.24; peer-reviewed by Y Sarikhani, S Das; comments to author 03.04.25; revised version received 14.04.25; accepted 21.04.25; published 29.05.25.

    Copyright

    ©Kyungmi Woo, Se Hee Min, Aeri Kim, Subin Choi, Gregory L Alexander, Terrence O’Malley, Maria D Moen, Maxim Topaz. Originally published in the Journal of Medical Internet Research (https://www.jmir.org), 29.05.2025.

    This is an open-access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in the Journal of Medical Internet Research (ISSN 1438-8871), is properly cited. The complete bibliographic information, a link to the original publication on https://www.jmir.org/, as well as this copyright and license information must be included.