This study used thematic analysis within an interpretivist framework, guided by the Standards for Reporting Qualitative Research [23,24]. To explore recurring patterns in participants’ experiences, semistructured interviews were conducted with individuals managing multiple chronic conditions, allowing for the identification of cross-cutting themes across diverse personal narratives. To investigate the concept of chronic illness management as a “career,” this study followed a multiphase qualitative approach designed to integrate both experiential depth and methodological comparison. The first phase involved the development of a semistructured interview guide, informed by literature review, expert consultation, and focus groups with patients. The second phase included refinement of the guide and data collection through interviews. In the third phase, interview transcripts were analyzed using two distinct methods: manual thematic analysis and artificial intelligence (AI)–assisted analysis using an LLM. The final phase focused on a comparative evaluation of the two analytical approaches, assessing similarities, differences, and unique contributions. This phased design allowed us to explore both patient experiences and the use of LLMs in qualitative analysis, as shown in Figure 1. This study followed an inductive analytical approach. While existing literature informed the development of the interview guide, no formal theoretical framework was systematically applied in either the manual or the LLM-assisted analysis.

This study was conducted in compliance with all relevant laws and institutional guidelines and was approved by the Tel Aviv University ethical review board. Formal ethical approval for the research was granted on February 13, 2023 (clinical trial number #0005962-1). Participation was voluntary, and informed consent was obtained from all participants before this study’s initiation. Participants were fully informed about the purpose, procedures, and potential risks of this study, and they retained the right to withdraw at any time without penalty. No financial or material compensation was provided. Confidentiality was ensured through anonymization of transcripts and secure data handling.

Table 1 summarizes the baseline characteristics of this study’s sample. The sample included 30 participants, 13 of whom were male, with an average age of 62.7 years (SD 7.3). The majority were either employed (n=19) or retired (n=10), with 7 of the 10 retirees attributing their retirement to medical reasons. The most common primary illnesses were diabetes (n=12) and rheumatic diseases (n=9). Most participants resided in the central region of the country (n=25), and 10 were the primary caregivers for a family member.

The traditional qualitative analysis identified six main themes (navigating daily life under chronic conditions, the role of support systems, health care and systemic challenges, emotional challenges and coping, self-management and adaptation, and sources of strength and motivation) as shown in Table 2, which presents the themes identified in the manual analysis along with illustrative participant quotes. The quotes presented in Table 2 provide contextual depth and support the interpretation of the identified themes. The LLM analysis also yielded six main themes (health management and adaptation, navigating health care systems, emotional challenges and coping, social and family dynamics, daily life and functional adjustments, and future planning and long-term concerns) as shown in Table 3, which presents the themes identified in the LLM-assisted analysis.

The traditional analysis identified 33 subthemes, while the LLM analysis uncovered 43 subthemes. Figure 2 presents a comparison between manual and LLM-assisted analyses, showing shared themes as well as themes identified uniquely by each approach. This figure visually summarizes the similarities and differences between the two analytical approaches. Many subthemes were consistent across both groups, with only 5 subthemes identified by the traditional analysis not appearing in the LLM analysis. Conversely, 1 subtheme was identified by the LLM analysis that was absent from the traditional analysis.

The comparative analysis revealed the following key findings identified by both methods. Illustrative quotes from participants appear in Tables 2 and 3.

This theme encompasses the strategies and efforts individuals use to manage and adapt to long-term health conditions. In the traditional analysis, illness management was described as an ongoing task embedded in daily routines. For example, patient 13 described actively managing his condition by monitoring and following up: “I don't give up on myself… I am writing to the doctor and ask for periodic tests. I have to keep an eye on my situation.” The LLM analysis reflected a similar focus, highlighting gradual acceptance and the integration of illness-related practices into one’s lifestyle. As patient 14 explained: “With awareness over the years, it really became part of my management… I learned to accept it.” More examples from participant narratives are presented in Tables 2 and 3.

This theme explores how family relationships and external support networks influence individuals’ ability to manage chronic illness. It appeared as the role of support systems in the traditional analysis and as social and family dynamics in the LLM model. Both analyses emphasized the significant role of family and community in shaping patients’ experiences, ranging from emotional encouragement and practical help to tensions or ambivalence in relationships. The findings reflect the diverse scope of support, from sincere care to feelings of burden or intrusion. For example, in the traditional analysis, patient 2 described strong family and community support: “The community in the neighborhood where I live was very helpful… I learned to appreciate the power of family.” Similarly, the LLM analysis also reflected the importance of support, as seen in patient 13’s statement: “Without such support from the environment it's impossible… without my family, it's impossible.”

This theme refers to the process of accessing and using health care services, and to the obstacles and frustrations patients face within the health care system. It was identified as health care and systemic challenges in the traditional analysis and as navigating health care systems in the LLM model. Participants described their journeys as bureaucratic, fragmented, and prolonged, often involving repeated referrals, inconsistent diagnoses, and delays before receiving effective care. These accounts reflect not only structural barriers but also a sense of exhaustion and uncertainty during the care-seeking process. For example, in the traditional analysis, patient 18 described the difficulty of receiving consistent diagnoses: “I went to about four dermatologists… each of them gave me a completely different diagnosis.” Similarly, the LLM analysis reflected the burden of repeated tests and long processes, as described by patient 20: “I had about a month filled with doing a million tests… I felt like I had been in the clinic for the whole month.”

This theme reflects the range of emotional responses individuals experience while managing chronic illness, including ongoing worry, frustration, guilt, and self-blame. Both traditional and LLM analyses highlighted how patients described emotional strain as part of daily life with chronic conditions. These expressions, while not always accompanied by explicit coping strategies, reveal an emotional labor patients engage in as they navigate uncertainty, treatment demands, and the psychological toll of illness. For example, in the traditional analysis, patient 6 described efforts to cope emotionally: “I try to protect my soul… with things that make me happy.” Similarly, the LLM analysis reflected emotional distress, as described by patient 14: “I had tears in my eyes… I feel like I failed.”

This theme captures the practical strategies and adjustments individuals make in their day-to-day lives to accommodate the demands of chronic illness. In the traditional analysis, it was labeled self-management and adaptation, while in the LLM model, it appeared as daily life and functional adjustments. Participants described changes to household routines, increased planning, and the gradual incorporation of external support when necessary. These narratives reflect a shift from full independence to negotiated interdependence, highlighting how people preserve autonomy while accepting limitations. For example, in the traditional analysis, patient 17 described changes in daily habits: “I have already changed my eating habits… I need to do some sport but I don't persist with it.” Similarly, the LLM analysis reflected adjustments in daily routines, as described by patient 2: “There are priorities… now with my eyes I'm not allowed to do many things so the helper comes twice a week.”

The comparative analysis revealed that certain themes emerged uniquely in one of the two analytic approaches: manual thematic analysis or LLM-assisted analysis. These differences reflect variation in how each method prioritized or interpreted aspects of participants’ narratives, rather than inconsistencies in accuracy.

This theme explores the factors that empower individuals to cope with challenges and pursue their goals. The traditional analysis uniquely identified this theme, highlighting both internal and external sources of motivation. Internal sources included personal faith, spiritual beliefs, and a proactive mindset that emphasized inner strength and optimism. External sources stemmed from caring responsibilities, encouragement from family members, and the need to maintain functional roles within the household.

These accounts illustrate how participants found motivation both within themselves and through their roles in the family and community. For example, patient 9 described a personal source of strength: “My strength comes from God… I believe in God.”

תחתית הטופס

The LLM analysis uniquely identified this theme, which reflects participants’ concerns about the future and the strategies they use to prepare for potential long-term challenges. These included considerations such as financial stability, medical planning, and personal life goals. Although such reflections appeared in the traditional analysis as well, they were embedded within a broader subtheme of emotional challenges and coping, rather than being treated as a distinct theme. The LLM analysis, however, grouped these expressions, emphasizing their frequency and framing them as a separate area of concern regarding future planning. For example, patient 18 described concerns about the future: “I think about the diseases that I have… it can shorten a person's life.”

This study compared traditional qualitative thematic analysis with AI-based analysis using LLM to explore patients’ experiences in managing chronic illness. The analysis revealed 5 shared themes: strategic approaches to chronic illness management, support systems and family dynamics, health care navigation and challenges, emotional challenges and coping, and routine adaptation in daily life. In addition, each method revealed unique themes: the traditional analysis identified sources of strength and motivation, encompassing subthemes such as personality, caregiving roles, faith, and a proactive mindset. The LLM analysis uniquely identified future planning and long-term concerns, including issues related to multiple health conditions. These findings highlight the complementary roles of traditional and LLM-assisted analysis and contribute to the emerging methodological literature on AI-assisted qualitative research, which explores how LLMs can support, extend, and challenge traditional approaches to qualitative analysis. These findings are discussed in more detail in the following sections, focusing on both shared and unique themes.

The overlap between themes identified by the traditional and LLM-based analyses suggests that both approaches capture core aspects of chronic illness management, consistent with existing literature on chronic care and patient-centered management [25-27]. This indicates that, despite differences in the analytical process, both methods provide a meaningful understanding of the patient’s experience and highlight their complementary strengths. For example, health care navigation and challenges were a recurring theme in both approaches, with patients describing the bureaucratic hurdles and frustrations of accessing care. Studies on health care navigation for chronic patients highlight several common issues, such as fragmented care, inconsistent communication among providers, and a lack of standardized care pathways [28]. These challenges often result in delays in receiving accurate diagnoses and appropriate treatments, as patients navigate a complex health care system [29]. Furthermore, patients with multiple chronic conditions often receive conflicting medical advice and face barriers such as long wait times and limited access to care [30-33]. Together, these findings suggest that both analytical approaches can capture key systemic challenges experienced by patients with chronic illness.

The importance of support systems and family dynamics across both approaches aligns with previous research showing the central role of social support in chronic illness management [34,35]. Participants described the emotional, logistical, and practical support provided by family members, friends, and community networks, which helped them maintain a sense of balance and resilience amid the demands of their chronic conditions. The support systems and family dynamics theme aligns with research emphasizing the importance of family caregivers and social support in improving health outcomes and quality of life for individuals with chronic illnesses [36]. Social support has been shown to provide not only practical assistance but also a crucial psychological buffer against the stress and uncertainty of managing chronic conditions [37]. By fostering a strong support system, patients can enhance their overall quality of life and better cope with the physical and emotional challenges of living with chronic illness [38]. However, while social support is invaluable, it also comes with its complexities. Family dynamics may introduce challenges such as caregiver burnout, role strain, or feelings of dependence among patients [39,40], highlighting the importance of developing interventions that support both patients and caregivers [41,42]. Together, these findings suggest that both analytical approaches capture not only the benefits of social support but also its challenges, reflecting the complexity of patients’ lived experiences.

The emotional challenges and coping impact of chronic illness were evident across both analytical approaches, highlighting the central role of emotional challenges in patients’ experiences. Participants described the ongoing struggle to maintain a sense of normalcy and control amid the uncertainty of living with chronic conditions, as well as the impact on their mental health, including feelings of depression, anxiety, and grief over the loss of their previous way of life [36,43]. These findings align with existing research demonstrating the psychological burden associated with chronic illness, including an increased risk of mental health issues such as depression and anxiety, which can exacerbate physical symptoms and hinder effective self-management [36,44,45]. To cope, patients described a range of strategies, such as cultivating a positive mindset, engaging in self-care, and seeking support from professionals and peer networks [46,47]. Addressing these emotional needs is critical, as interventions targeting mental health can significantly improve overall health outcomes [48]. Together, these findings suggest that both analytical approaches capture not only the emotional burden of chronic illness but also the coping strategies patients use to manage these challenges.

Finally, chronic illness management, as reflected in both analytical approaches, encompasses a wide range of practices and experiences that can be viewed through both positive and challenging lenses. On the one hand, disease management can serve as a source of education and empowerment, enabling patients to take an active role in their care through self-monitoring and lifestyle adjustments, which may improve autonomy and health outcomes [49]. On the other hand, the burden of chronic illness management can be overwhelming, leading to burnout, reduced quality of life, and ongoing challenges in navigating the health care system [50]. Together, these findings highlight the complex and sometimes conflicting nature of chronic illness management, as experienced by patients. However, both analyses focused on identifying themes rather than applying a specific theoretical framework, and therefore remain mainly descriptive.

The fact that the traditional analysis uniquely identified the theme of sources of strength and motivation suggests its strength in capturing more nuanced and deeply personal aspects of patient experience. This theme encompasses subthemes such as personality as a driver of action, taking care of others, faith as a source of strength, and a proactive approach. Participants described how personal beliefs, a sense of purpose, and relationships influenced their ability to manage their condition, such as patient 15’s connection to their professional identity. These findings are consistent with previous research showing that resilience, self-efficacy, and spiritual beliefs can play a central role in chronic illness management and are associated with improved coping and health outcomes [51-53]. Taking care of others can also strengthen patients’ ability to navigate care and maintain meaningful roles [54]. In addition, a positive mindset has been associated with better self-care behaviors and disease outcomes. These findings also reflect the role of personality traits, such as resilience and self-efficacy, in driving active engagement in illness management, supporting the subtheme of personality as a driver of action [52,55]. The subtheme of a proactive approach further reflects patients’ expectations for more active and supportive care. Participants emphasized the need for health care providers to take a more active role, such as scheduling follow-up appointments and maintaining ongoing engagement, rather than relying on patients to initiate contact. This finding aligns with existing literature on chronic disease management, which emphasizes the importance of planned and coordinated care that anticipates and addresses patient needs [49,56]. At the same time, this approach may not suit all patients, as some may perceive it as limiting their autonomy or creating additional pressure, highlighting the need to balance proactive care with respect for patients’ preferences and independence [57]. Together, these findings suggest that the traditional analysis can capture not only patient experiences, but also deeper, context-dependent aspects such as personal motivations, values, and expectations regarding care, which may be less evident in more structured analytical approaches.

The identification of multiple health conditions as a distinct subtheme highlights the ability of the LLM-assisted analysis to capture the complexity and accumulation of challenges faced by patients managing more than one chronic illness. This subtheme reflects the combined burden of multiple conditions, each with its own symptoms, treatments, and impact on daily life. As described by patient 7, “It's not just one thing, it's everything together. I have diabetes, high blood pressure, arthritis, and now this new thing with my heart. It's just too much sometimes.” Managing multiple chronic illnesses often requires navigating a fragmented health care system, coordinating care among multiple providers, and following complex and sometimes conflicting treatment plans [58]. Patients may also experience a compounding effect, where one condition exacerbates the symptoms or management challenges of another, increasing the overall burden and complexity of care [59]. These findings suggest that LLM-assisted analysis may be particularly effective in identifying patterns related to complexity, accumulation, and co-occurring conditions within patient narratives. By highlighting these patterns, this approach may support a more system-level understanding of chronic illness management.

Overall, this study highlights the complementary value of traditional and LLM-assisted qualitative analysis in understanding chronic illness management. While traditional analysis provided deeper insight into personal meanings, motivations, and patient perspectives, the LLM-assisted approach was effective in identifying broader patterns and capturing the complexity of managing multiple conditions. Together, these findings suggest that integrating both approaches can enhance the depth, scope, and applicability of qualitative health research, particularly in complex clinical contexts.

This study provides a valuable comparison between traditional qualitative analysis and AI-based analysis using Claude 3.5 Sonnet. However, this reliance on a single LLM constitutes a key limitation, as it does not account for potential variations in performance and insights that may arise from using other LLMs, such as ChatGPT, Gemini (Google LLC), or other emerging systems. Future studies should explore a comparative evaluation across multiple LLMs to better understand the range of capabilities, biases, and thematic coverage. Another limitation relates to the dataset and its contextual dependency. While the dataset was sufficient for this analysis, its characteristics may have influenced the themes identified by both methods. Broader and more diverse datasets could support more comprehensive comparisons. Participant recruitment through primary care physicians may have introduced selection bias, as individuals who are more communicative or more engaged with the health care system may have been more likely to be referred to. This may have influenced the range of perspectives represented in the sample. The comparison between the two approaches was based on predefined qualitative criteria and remained primarily interpretive and descriptive, without formal quantitative evaluation. In addition, the traditional qualitative analysis relied on a single researcher, which may have increased the risk of interpretive bias, although reflexive practices were used to mitigate this. Another limitation is that both the manual and LLM analyses were not theory-driven, resulting in findings that are primarily descriptive rather than theoretically grounded. Future research should explore how theoretical frameworks can be integrated into both manual and LLM-assisted analysis. Similarly, while LLMs offer scalability and efficiency, their output depends on the training data, which introduces biases or overlooks important contextual nuances. Moreover, the use of an LLM raises ethical considerations, including transparency, interpretability, and the potential for misrepresentation of participants’ voices. In this study, the model operated solely on anonymized transcripts and was used to support, rather than replace, human interpretation. Finally, this study did not assess the reliability of LLM-generated themes or their clinical applicability to real-world scenarios. Therefore, these findings should be interpreted as exploratory thematic observations rather than clinically validated psychosocial classifications. Future research should integrate mixed-method approaches to evaluate the validity, reproducibility, and practical relevance of themes generated by LLMs. More broadly, future studies should aim to integrate AI-driven tools into qualitative research designs in ways that enhance both efficiency and analytical depth. The development of hybrid frameworks that use LLM-based pattern recognition with human-led contextual interpretation represents a promising direction. Additionally, longitudinal research and real-world applications will be important for assessing the stability and clinical relevance of LLM-generated insights.

This study examined the use of traditional qualitative thematic analysis and LLM-based analysis to explore patients’ experiences in managing chronic illness. Both approaches identified similar core themes, while also revealing distinct insights.

These findings suggest that the two approaches offer complementary perspectives. Traditional analysis enabled a rich, interpretive understanding of patient narratives, while the LLM approach supported efficient and systematic identification of thematic structures. While AI-based methods hold potential to support qualitative research—particularly in scaling and organizing data—they require careful validation to ensure cultural and contextual sensitivity. Taken together, this study highlights the value of a hybrid approach, integrating both human insight and AI capabilities, to enhance the depth and scope of qualitative health research.