Impact of hospital complaint handling on promoting high-quality development of hospitals via an emotional language analysis model: a case study of a tertiary hospital service center in Quanzhou city, Fujian province

Changxi Zheng, Yi Zhang, Xiaolong Lian, Jinxiu Ke, Hongxia Chen, Yinwen ChenFrontiers in Health Services 2025 · DOI

This study, published in Frontiers in Health Services, presents a data-driven framework for hospital complaint management that combines psychological profiling with advanced sentiment analysis. By integrating the DISC behavioral model with an improved clustering algorithm (KANN-DBSCAN), researchers demonstrated that tailoring communication strategies to a patient's specific behavioral style can significantly boost satisfaction rates and reduce the financial burden of medical disputes.

TL;DR

Researchers developed a hybrid model that uses the DISC behavioral framework and an optimized clustering algorithm to analyze and respond to patient complaints. The intervention increased patient satisfaction from 83.2% to 93.4% and significantly reduced the total amount of financial compensation paid out by the hospital.

What Problem Does This Paper Address?

In the healthcare industry, patient complaints are often viewed as a "necessary evil," but they are actually critical metrics for service quality. However, most hospitals handle complaints using traditional, manual methods that rely on the subjective judgment of staff. These methods often fail to capture the underlying emotional needs of the patient, leading to escalated conflicts, legal petitions, and high financial settlements.

The challenge lies in the "unstructured" nature of complaint data—free-text comments filled with nuance, frustration, and varying expectations. This paper seeks to bridge the gap between raw data and actionable management by providing a systematic way to categorize complaints and, more importantly, a psychological roadmap for how to resolve them.

Methods

The researchers employed a multi-stage analytical pipeline:

  1. Data Preprocessing: Raw complaint text was cleaned using mechanical compression, word segmentation, and stop-word filtering.
  2. Sentiment Analysis: Using the Boson sentiment lexicon, the team assigned scores to patient comments to quantify the intensity of negative emotions.
  3. The DISC Behavioral Model: This is the core "human" element of the study. Complainants were categorized into four behavioral types:
  4. D (Dominance): Seek control; need professional, immediate solutions.
  5. I (Inducement): Value social interaction; need to feel heard and respected.
  6. S (Submission/Steady): Avoid conflict; need patience and a safe space.
  7. C (Compliance): Fact-oriented; need logical, well-structured explanations.
  8. KANN-DBSCAN Clustering: Standard DBSCAN (Density-Based Spatial Clustering of Applications with Noise) is powerful but sensitive to manual parameter settings. The authors used KANN-DBSCAN, which uses a K-nearest neighbor approach to automatically determine the optimal "neighborhood" size ($\epsilon$) and density threshold ($P_{min}$), making the clustering of complaint themes more precise and less prone to human error.

Datasets

The study utilized a dataset of 1,278 hospital complaints from a tertiary teaching hospital in Quanzhou, Fujian Province, China, spanning January 2022 to December 2024.

  • Control Group (n=334): Complaints from 2022–2023 handled via standard procedures.
  • Intervention Group (n=341): Complaints from 2024 handled using the DISC-informed strategy.
  • Lexicon: The publicly available Boson sentiment lexicon was used for scoring.

Key Insights and Ideas

  • Psychology + Data Science: The most significant insight is that sentiment analysis alone isn't enough to resolve a conflict. Knowing a patient is "angry" (sentiment) is less useful than knowing they are a "C-type" personality who requires a factual, logical explanation to be appeased.
  • Evolution of Expectations: The clustering analysis revealed a shift in patient concerns. In 2023, complaints centered on "poor attitude" and "environment." By 2024, after basic service improvements, complaints shifted toward "medical efficiency" and "communication clarity." This suggests that as a hospital's "soft" services improve, patients' expectations for "hard" technical efficiency increase.
  • Reducing "Defensive" Payouts: The study found that while the rate of compensation didn't change drastically, the total amount paid out dropped significantly. This suggests that better communication can resolve the emotional "hurt" that often drives patients to demand higher financial settlements.

Key Results

  • Satisfaction Boost: Overall satisfaction in the intervention group was 93.39%, compared to 83.24% in the control group.
  • Statistical Significance: Multivariable logistic regression showed that the DISC-based intervention was independently associated with higher satisfaction, with an Adjusted Odds Ratio (aOR) of 3.06.
  • Emotional De-escalation: The "negative emotion score" derived from the sentiment analysis dropped from -17 in 2023 to -8 in 2024, indicating a measurable improvement in the hospital's service climate.
  • Efficiency: The KANN-DBSCAN model successfully identified specific service deficiencies (e.g., "lack of timeliness," "failure to explain clearly") that allowed management to implement targeted staff training.

Limitations and Future Work

The authors acknowledge several hurdles:

  • Clustering Sensitivity: While KANN-DBSCAN is an improvement, it still struggles with overlapping clusters where high-density and low-density regions meet.
  • Interpretability: The model is excellent at grouping data, but it lacks "Explainable AI" (XAI) features that could tell a manager why a specific complaint was clustered a certain way.
  • Human Factor: The success of the DISC model relies heavily on staff training. If frontline workers cannot accurately "read" a patient's behavioral type under stress, the model's effectiveness drops.
  • Future Direction: The authors suggest developing a closed-loop, real-time system that can analyze a complaint as it is being filed and immediately recommend a response protocol to the staff member on duty.

Summary updated

Abstract

By analyzing patients' psychological needs and expectations, this study aims to minimize financial compensation and reduce patient appeals while enhancing overall complaint resolution satisfaction, which provides medical institutions with a more comprehensive, effective, and personalized complaint-handling strategy while simultaneously improving patients' healthcare experiences.
BibTeX
@inproceedings{zheng2025impact,
  author = {Changxi Zheng and Yi Zhang and Xiaolong Lian and Jinxiu Ke and Hongxia Chen and Yinwen Chen},
  title = {{Impact of hospital complaint handling on promoting high-quality development of hospitals via an emotional language analysis model: a case study of a tertiary hospital service center in Quanzhou city, Fujian province}},
  booktitle = {Frontiers in Health Services},
  year = {2025},
  doi = {10.3389/frhs.2025.1610004},
}
Figure