H. M. Lian
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
H. M. Lian has a total of 8 co-authors. Here are the top co-authors by number of shared publications:
- Jiao Long
Affiliated Hospital of North Sichuan Medical College, Nanchong City, China
1 shared publication - Mengjiao Yang
Affiliated Hospital of North Sichuan Medical College Nanchong Sichuan China
1 shared publication W. J. Wang
Affiliated Hospital of North Sichuan Medical College Nanchong Sichuan China
1 shared publication- Y. L. Wang
Affiliated Hospital of North Sichuan Medical College Nanchong Sichuan China
1 shared publication Y. Liu
Affiliated Hospital of North Sichuan Medical College Nanchong Sichuan China
1 shared publicationY. R. Zhang
Affiliated Hospital of North Sichuan Medical College Nanchong Sichuan China
1 shared publication
Latest Publications
- Open Access
- 5 November 2026
- Published by Wiley in Journal of Clinical Nursing
Co‐Occurrence Network of Postoperative Symptoms Following Heart Valve Replacement: An Ising Model Analysis of Patient Self‐Reports From the Zhihu Platform
Background: Online patient narratives may complement structured assessments of symptoms following heart valve replacement. Network analysis can describe conditional associations between symptom mentions, but its clinical interpretation depends on extraction validity and estimation stability. Aim: To describe symptom mention proportions, conditional associations, centrality and community structure in Zhihu narratives concerning recovery after heart valve replacement. Methods: This retrospective cross‐sectional text‐mining study screened 5040 entries and identified 1017 texts classified as patient self‐reports. Of these, 421 contained at least one mention from a 38‐item symptom dictionary. Fifteen items met the network inclusion criteria. An Ising model was estimated using L1‐regularized logistic regressions, EBIC γ = 0.25 and the AND rule. Edge uncertainty and centrality stability were assessed using 1000 nonparametric and 1000 case‐dropping bootstrap replications, respectively. Communities were explored using Leiden modularity optimization. Extraction was validated against adjudicated independent annotations of 38 items in 150 posts randomly sampled from the 1017 posts. Results: Generalized pain (34.20%), anxiety (22.80%) and fatigue (16.39%) were the most frequently mentioned symptoms. The estimated network retained six positive edges and seven isolated nodes. Dyspnea–chest tightness was the strongest edge (2.43; 95% bootstrap interval 1.18–3.26); the intervals for the other five retained edges included zero. Dyspnea had the highest strength and one‐step expected influence (both 5.79). The CS(cor = 0.7) coefficient was 0.283 for both indices. Leiden partitioning identified three multi‐node communities and seven singleton communities. Automated extraction achieved precision of 97.35%, recall of 81.33% and an F1‐score of 0.886. Conclusion: The symptom‐mention network was sparse, with dyspnea showing the highest estimated centrality. Frequent mention did not imply high network centrality. Limited centrality stability and residual extraction errors preclude firm identification of clinical intervention targets. External validation of extraction performance and prospective clinical validation are required. Relevance to Clinical Practice: Generalized pain, anxiety and fatigue were frequently mentioned in the included narratives, whereas dyspnea had the highest estimated network centrality. These different findings may help frame questions for comprehensive postoperative symptom assessment, but neither mention frequency nor centrality establishes a treatment priority. The results do not demonstrate that intervening on a particular symptom improves recovery, and they require validation in clinically characterized cohorts before informing symptom management.
- 5 November 2026
- Published by Wiley in Journal of Clinical Nursing
Open AccessCo‐Occurrence Network of Postoperative Symptoms Following Heart Valve Replacement: An Ising Model Analysis of Patient Self‐Reports From the Zhihu Platform
Background: Online patient narratives may complement structured assessments of symptoms following heart valve replacement. Network analysis can describe conditional associations between symptom mentions, but its clinical interpretation depends on extraction validity and estimation stability. Aim: To describe symptom mention proportions, conditional associations, centrality and community structure in Zhihu narratives concerning recovery after heart valve replacement. Methods: This retrospective cross‐sectional text‐mining study screened 5040 entries and identified 1017 texts classified as patient self‐reports. Of these, 421 contained at least one mention from a 38‐item symptom dictionary. Fifteen items met the network inclusion criteria. An Ising model was estimated using L1‐regularized logistic regressions, EBIC γ = 0.25 and the AND rule. Edge uncertainty and centrality stability were assessed using 1000 nonparametric and 1000 case‐dropping bootstrap replications, respectively. Communities were explored using Leiden modularity optimization. Extraction was validated against adjudicated independent annotations of 38 items in 150 posts randomly sampled from the 1017 posts. Results: Generalized pain (34.20%), anxiety (22.80%) and fatigue (16.39%) were the most frequently mentioned symptoms. The estimated network retained six positive edges and seven isolated nodes. Dyspnea–chest tightness was the strongest edge (2.43; 95% bootstrap interval 1.18–3.26); the intervals for the other five retained edges included zero. Dyspnea had the highest strength and one‐step expected influence (both 5.79). The CS(cor = 0.7) coefficient was 0.283 for both indices. Leiden partitioning identified three multi‐node communities and seven singleton communities. Automated extraction achieved precision of 97.35%, recall of 81.33% and an F1‐score of 0.886. Conclusion: The symptom‐mention network was sparse, with dyspnea showing the highest estimated centrality. Frequent mention did not imply high network centrality. Limited centrality stability and residual extraction errors preclude firm identification of clinical intervention targets. External validation of extraction performance and prospective clinical validation are required. Relevance to Clinical Practice: Generalized pain, anxiety and fatigue were frequently mentioned in the included narratives, whereas dyspnea had the highest estimated network centrality. These different findings may help frame questions for comprehensive postoperative symptom assessment, but neither mention frequency nor centrality establishes a treatment priority. The results do not demonstrate that intervening on a particular symptom improves recovery, and they require validation in clinically characterized cohorts before informing symptom management.