Haematological malignancies (HMs) are a heterogeneous group of aggressive neoplasms originating from the haematopoietic system, including leukaemia, lymphoma and multiple myeloma. They account for a meaningful share of global cancer incidence and mortality and pose substantial treatment and survivorship challenges. Although advances in targeted therapy and immunotherapy have extended survival, HM patients continue to face long treatment cycles, high drug costs and psychological burdens. In this context, proactive self-care and health behaviour abilities — the self-rated capability to implement actions such as medication adherence, symptom monitoring, nutritional management and exercise — are important determinants of quality of life and recovery.
This study applied a person-centred latent profile analysis (LPA) approach to identify distinct patterns of health behaviour abilities among hospitalised patients with HMs and to explore associated influencing factors across multiple ecological levels. The goal was to characterise population heterogeneity and inform targeted clinical interventions to enhance patient self-management.
This work used a cross-sectional design conducted between October and December 2024. Convenience sampling recruited inpatients from haematology-oncology departments at three provincial hospitals in Hunan Province, China. The study was approved and administered with hospital nursing department consent; patients completed anonymous questionnaires after being informed about study purpose and procedures.
Eligible participants were adults (≥18 years) with a pathological diagnosis of an HM, currently hospitalised and receiving treatment, and able to communicate and complete the survey. Exclusion criteria included cognitive impairment, diagnosed mental disorders, or comorbid diseases of vital organs (heart, brain, kidneys). A total of 500 questionnaires were distributed; 492 were returned and, after quality checks and exclusions, 486 valid responses remained (effective recovery rate 98.78%). Approximately half (49.6%) of respondents were aged 40–59 years.
Measured variables were structured using the Social Ecological Model into intrapersonal, microsystem, exosystem and macrosystem levels. Instruments included:
The study also collected 14 demographic and clinical items including age, gender, marital status, education, occupation, residence, monthly family income, diagnosis, chronic diseases, disease course, treatment stage, relapse, insurance and complications.
Two trained research assistants and researchers oversaw data collection. Participants completed questionnaires independently and anonymously; investigators assisted orally for those with difficulty. Completed questionnaires were checked on site for errors or omissions and corrected immediately. After double-person verification, invalid questionnaires were removed based on response patterns and completion time criteria.
LPA was performed in Mplus v8.0 using the four dimension scores of the health behaviour abilities scale as observed indicators. Models with one to four profiles were compared using AIC, BIC, adjusted BIC (smaller is better), entropy (closer to 1 indicates higher classification accuracy), and likelihood-based tests (LMR and BLRT). Model interpretability guided final selection. SPSS v28.0 was used for descriptive analyses, χ2 or rank-sum tests to compare groups, and multivariable multinomial logistic regression to identify predictors of profile membership. A two-tailed p<0.05 was considered statistically significant. Sample size met events-per-variable guidance for multivariable regression given 18 candidate independent variables and an achieved sample of 486.
A total of 486 valid questionnaires were analysed. The mean overall score for health behaviour abilities was 47.96 (SD 23.29) on the 0–112 scale, indicating a generally low level of self-rated ability across the cohort. The study reports mean scores for the four subdimensions but these specific subscale means are not reproduced here because the source truncated the results section in the provided text.
Latent profile analysis identified three distinct profiles reflecting heterogeneous patterns across the four health-behaviour dimensions. The three labelled profiles were:
“Exercise-weak survival profile” — representing 41% of participants and characterised by relative weakness in exercise alongside survival-oriented coping patterns.
“Exercise-deficient developmental profile” — representing 45% of participants and marked by deficits in exercise with distinct developmental features in other domains.
“Balanced-robust profile” — representing 14% of participants and showing relatively balanced and robust abilities across nutrition, exercise, psychological adjustment and health responsibility.
These profiles demonstrate substantial heterogeneity in self-rated health behaviour abilities within the HM inpatient population.
Multivariable multinomial logistic regression identified several factors associated with membership in the latent profiles. Significant influencing factors included age, educational level, social support, financial toxicity and psychological distress. The source reports these variables as determinants of profile assignment but specific odds ratios, confidence intervals and p values were not provided in the extracted text and thus are not reproduced here.
Strengths of the study include the application of latent profile analysis combined with multinomial logistic regression as a person-centred methodological approach, and sampling across multiple primary HM diagnoses to improve representativeness. Limitations include reliance on online self-reported questionnaires, which introduces response and recall biases; the cross-sectional design, which precludes causal inference; and recruitment limited to hospitals within a single city, which may restrict generalisability.
Among hospitalised patients with HMs in this sample, overall health behaviour abilities were low and exhibited three distinct latent profiles. The study suggests that clinical healthcare providers should implement targeted, profile-specific interventions to strengthen patients’ self-management capabilities, such as addressing exercise deficits and tailoring support according to age, education, social support, financial toxicity and psychological distress levels.
The data supporting this study’s findings are available from the corresponding author on reasonable request, consistent with the source article’s statement.