Near-Suicide Phenomenon: An Investigation into the Psychology of Patients with Serious Illnesses Withdrawing from Treatment

The study introduces the concept of "near-suicide," where patients with serious illness or injury choose to withdraw from medical treatment to avoid financially devastating their families, thereby risking death without direct self-harm. In Vietnam’s context of significant out-of-pocket healthcare costs and partial insurance coverage, such decisions may be driven by subjective cost-benefit evaluations rooted in personal and familial values. The research question: How do the seriousness of the patient's illness or injury and the subjective evaluation of the patient's and family's financial situation after paying treatment fees affect the final decision on the treatment process? The purpose is to explore this extreme decision-making phenomenon using an information-processing framework (mindsponge) combined with Bayesian analytics, offering insights for evidence-based health policy to reduce such outcomes.

The paper reviews multiple perspectives on suicide, including classical and contemporary theories: Durkheim’s typology (egoistic, altruistic, anomic, fatalistic), the Economic Theory of Suicide, the Three-Step Theory (3ST), the Interpersonal Theory of Suicide (ITS), the Integrated Motivational-Volitional model (IMV), and Fluid Vulnerability Theory (FVT). It discusses biological, psychological, and social determinants and highlights ethical and philosophical debates (e.g., Mill, Bentham, Szasz, Confucian viewpoints). The authors argue that existing models, largely developed in Western contexts, may not fully explain context-specific, collectivist-driven phenomena like "near-suicide" in Vietnam. The data problem in suicide research is emphasized (difficulties in collecting data on completed suicides and extreme events), motivating an indirect approach using mindsponge reasoning to study extreme psychosocial phenomena via more accessible health data.

Design and data: Secondary dataset of 1042 patients collected from major hospitals in Northern Vietnam (e.g., Viet Duc Hospital, Bach Mai Hospital in Hanoi; Viet Tiep and Kien An in Haiphong; Uong Bi in Quang Ninh). Data collection proceeded in three phases from 2014 to March 2016, using in-depth interviews with patients or family members, including sensitive socioeconomic and financial questions. Ethical standards (ICMJE, WMA Declaration of Helsinki, Vietnam MOH Decision 460/QD-BYT) were followed; data and protocol were validated and published in data journals (Data in Brief, Data). Participants provided informed consent. Key sample characteristics: mean age 45.4 years; most had high school education; illness seriousness predominantly bad (49.9%) or emergency (27.4%); ~70% had valid insurance; 109 patients stopped mid-treatment (4.5%) or quit early (6.0%). Variables: End (binary treatment outcome: 1 = recovered/need follow-up; 0 = stopped in middle/quit early), Illness (categorical: 1 Ill/light, 2 bad, 3 emergency), Burden (financial burden post-fee payment: 1 no adverse effect, 2 affected but not worrying, 3 seriously affected, 4 destitute/bankrupt; treated as numerical). Model: End ∼ α[Illness] + Burden (logit). Analytical approach: Bayesian Mindsponge Framework (BMF) analytics, integrating mindsponge reasoning for conceptualization and Bayesian inference with MCMC for estimation. Software: bayesvl R package. Estimation details: four chains, 5000 iterations each with 2000 warm-up. Model validation: PSIS-LOO (all k < 0.5 indicating adequate fit), convergence diagnostics (n_eff > 1000, Rhat = 1 across parameters; confirmed via trace, Gelman-Rubin-Brook, and autocorrelation plots). Priors: primary estimation used uninformative Normal(0,10) priors; robustness checked via informative prior for Burden Normal(0,0.5), yielding consistent patterns (prior-tweaking robustness).

- Higher perceived financial burden significantly reduced the probability of recovering/continuing treatment and increased the likelihood of stopping/quitting (near-suicide decision). Burden coefficient mean −1.33 (SD 0.18) under uninformative priors; posterior mass entirely negative. - Greater illness seriousness also reduced the probability of continuing treatment, with intercepts decreasing from Ill/light to Bad to Emergency; Emergency category had the lowest intercept posterior distribution, confirming a strong negative gradient with seriousness. - Probabilistic scenarios: Among patients with the most serious health issues who perceived continuing treatment would lead to destitution/bankruptcy, only about 24% were estimated to choose to continue treatment. In contrast, among seriously ill patients who believed paying fees would not adversely affect finances, about 95% were estimated to continue. For light illness, the probability to continue exceeded 90% regardless of financial status. - Model adequacy and convergence were supported: PSIS-LOO k-values < 0.5; n_eff > 1000 and Rhat = 1 for parameters under both prior settings, indicating robust inference.

Findings indicate that near-suicide decisions arise from subjective, context-sensitive cost-benefit evaluations where patients weigh personal survival and suffering against the financial well-being and future of their families. This aligns with perspectives treating suicide-related decisions as rational, though emotionally and culturally embedded, choices rather than purely pathological processes. In Vietnam’s collectivist context influenced by Confucianism, Taoism, and Buddhism, family-centered values can intensify the perceived benefits of withdrawing from treatment to avoid familial destitution. Additional sociocultural pathways (e.g., hope in traditional remedies or moral luck) may shape perceived benefits. The phenomenon reflects altruistic motives (in Durkheim’s sense) and “biological altruism,” prioritizing family welfare. Comparatively, near-suicide differs from euthanasia in means yet similarly reflects choosing the subjectively least-worst option. Community coping (e.g., co-located patient support groups) emerges as informal strategies. Methodologically, the mindsponge framework combined with BMF analytics demonstrates utility in studying extreme psychosocial phenomena when direct data are difficult or ethically constrained, enabling parsimonious, transparent, and robust inference. Policy implications include reducing financial barriers (improving effective coverage rather than only expanding nominal universal coverage) to mitigate near-suicide risks, targeting disadvantaged non-resident poor patients, and pursuing evidence-based adjustments to health insurance design and implementation.

Using Bayesian analysis of data from 1042 Vietnamese patients, the study shows that more severe illness and higher perceived financial burden markedly increase the likelihood of withdrawing from treatment (near-suicide). Only about one-quarter of the most severely ill patients would continue treatment if they believed doing so would push their families into destitution, whereas continuation probabilities were about 95% for similarly ill patients perceiving no adverse financial impact. These findings substantiate the near-suicide phenomenon in a collectivist, resource-constrained context and underscore the need for evidence-based health policies to reduce financial burdens and prevent fatal treatment-withdrawal decisions. The study also illustrates the effectiveness of mindsponge-based reasoning and BMF analytics for investigating extreme psychosocial phenomena.

- Generalizability: Data are from Vietnamese patients, primarily in Northern Vietnam, and findings may be influenced by local sociocultural contexts; patterns may vary elsewhere. - Temporal/contextual factors: Patient perceptions and behaviors might have been affected by broader events (e.g., COVID-19) not explicitly modeled. - Ethical oversight: While data and protocols were validated and the dataset is open under CC BY 4.0 with adherence to international ethical standards, no institutional review board approval was available in Vietnam at the time; thus, formal institutional ethics review was not obtained. - Measurement and design: The study relies on subjective evaluations of financial burden and illness seriousness; cross-sectional observational design limits causal inference.