In a multicenter cross-sectional study of 3,429 hospital nurses across Anhui Province, China, investigators found that higher self-reported humanistic caring knowledge was positively associated with reported caring practice, but whether this association varied by perceived departmental caring climate depended on model specification. Analyzing survey data collected between April 5 and May 30, 2025, across eighty-five reconciled facility groups, the team examined whether a supportive workplace atmosphere amplifies the translation of caring knowledge into practice. In their adjusted additive linear model, every one-point higher self-reported knowledge score was associated with a 0.221-point higher reported practice score (95 percent confidence interval, 0.183 to 0.259). However, when evaluating whether this slope differed between nurses rating climate as very good versus fair or poorer, the primary linear model demonstrated a difference of only 0.005 points (95 percent confidence interval, negative 0.094 to positive 0.104; p = 0.917), failing to demonstrate linear moderation by perceived workplace climate.
The conceptual motivation for this research stems from broader inquiries into how nursing work environments and emotional engagement intersect with health service quality. Humanistic nursing care highlights core interpersonal dimensions of patient treatment, such as respecting personal dignity, attending to emotional needs, and maintaining clear communication. While longitudinal studies have linked improved working conditions to favorable patient experiences and lower burnout, prior cross-sectional surveys often assumed that acquiring knowledge through education automatically translates into routine bedside practice. The authors noted that a supportive departmental setting might provide more practical opportunities to apply caring knowledge during clinical shifts. To examine whether the knowledge-to-practice association genuinely varies with perceived climate, the researchers designed an analytical protocol to separate linear trends from potential non-linear dynamics, while properly accounting for institutional clustering across regional hospitals.
Recruitment initially targeted nurses across five cities in Anhui Province—Hefei, Fuyang, Bengbu, Chuzhou, and Xuancheng—selecting one or two tertiary Grade A hospitals per city and inviting all nursing staff. However, returned questionnaires expanded beyond this framework, incorporating secondary hospitals alongside identified community and township facilities. Out of 3,488 submitted questionnaires, forty-five were excluded due to unresolved facility identities and fourteen lacked valid age data, leaving an analytical sample of 3,429 nurses across ninety-three operational sites within eighty-five reconciled facility groups. The median group size was forty-three participants, ranging from one to 133 nurses per group. Among participants, 3,297 were female (96.2 percent) and 132 were male (3.8 percent). The mean participant age was 34.86 years (standard deviation, 7.03), and median work experience reached 12.00 years. Because selection probabilities were unavailable, the authors emphasized that the sample cannot be viewed as a probability sample of the provincial workforce.
Educational backgrounds and clinical roles covered varied practice environments across the regional health system. Bachelor’s degrees were held by 85.3 percent of respondents, junior college diplomas or lower qualifications by 13.7 percent, and graduate degrees by 1.0 percent. Administrative management roles were held by 18.5 percent of participants, while 81.5 percent occupied non-administrative bedside positions. Departmental distributions included internal medicine with 29.6 percent, surgery with 20.5 percent, emergency departments with 6.1 percent, pediatrics with 5.8 percent, obstetrics and gynecology with 5.1 percent, intensive care with 4.9 percent, outpatient units with 2.8 percent, and other departments accounting for 25.3 percent. Average monthly night shifts over the preceding year were reported as zero by 13.8 percent of nurses, one to five shifts by 27.5 percent, six to ten shifts by 32.8 percent, eleven to fifteen shifts by 10.6 percent, and sixteen or more shifts by 15.3 percent. Additionally, 68.1 percent reported prior caring-related training, and 85.7 percent reported voluntarily choosing nursing.
The survey quantified caring dimensions using eighteen self-reported knowledge items, twenty-three reported practice items, and twenty-two attitude items, all answered on a five-point agreement scale coded from one for strongly disagree to five for strongly agree. Individual domain scores represented unweighted arithmetic item means, reflecting self-assessed agreement rather than objective knowledge testing or direct behavioral observations. Self-reported caring knowledge yielded an overall mean of 3.77 (standard deviation, 0.76), while reported practice showed a higher mean of 4.47 (standard deviation, 0.54). Departmental caring climate was evaluated using a single item categorized into very good for 46.0 percent of nurses, good for 37.4 percent, and fair or poorer for 16.6 percent. Notably, 23.9 percent of respondents achieved the maximum score of 5.00 on the practice scale. This ceiling concentration was unevenly distributed across perceived climates, reaching 34.3 percent in the very good group compared with 15.3 percent in the good group and 14.2 percent in the fair or poorer group.
To account for institutional clustering across the eighty-five facility groups, researchers utilized linear regression models with facility fixed effects and bias-reduced cluster-robust covariance corrections. Models adjusted for age, sex, education, administrative role, department, night shifts, previous training, voluntary nursing choice, and survey month. In the baseline specification without individual covariates, a one-point increase in knowledge was associated with a 0.258-point higher practice score. Adjusting for all covariates yielded a common knowledge slope of 0.221 (95 percent confidence interval, 0.183 to 0.259). When estimating climate-specific slopes, the knowledge slope was 0.225 for very good, 0.215 for good, and 0.220 for fair or poorer climates. The primary contrast comparing very good to fair or poorer categories produced a slope difference of 0.005 (95 percent confidence interval, negative 0.094 to positive 0.104; p = 0.917). A joint interaction test yielded an F-statistic of 0.036 with degrees of freedom of 2 and 45.67 (p = 0.965), confirming the absence of linear slope heterogeneity across climate categories.
While linear models showed no slope variation, non-linear sensitivity analyses revealed that climate moderation findings depended on functional form assumptions. When modeling knowledge with a two-degree-of-freedom natural cubic spline, average slopes over the central knowledge interval of 3.444 to 4.111 were 0.231 in very good, 0.283 in good, and 0.343 in fair or poorer climates. Under this spline specification, the very good minus fair or poorer contrast was negative 0.112 (95 percent confidence interval, negative 0.182 to negative 0.042; p = 0.003), with a joint interaction test of F = 4.921 (degrees of freedom, 4 and 43.05; p = 0.002). All average slopes remained positive, meaning the negative contrast indicated a flatter rate of increase in the very good group over this specific range. The authors noted that this pattern likely reflects practice-score ceiling concentration, as 34.3 percent of nurses in very good climates reached the maximum score. Conversely, a bounded fractional-response logit model estimated an original-scale average-slope contrast of 0.019 (95 percent confidence interval, negative 0.056 to positive 0.093; p = 0.614), demonstrating how non-linear models alter empirical conclusions.
Sensitivity checks focusing on instrument content and facility linkage confirmed the stability of the linear results. To address potential conceptual overlap between climate perceptions and practice questions involving administrative systems, researchers analyzed a twenty-item practice composite omitting training participation, departmental workflows, and multidisciplinary caring teams. This restricted scale yielded a mean score of 4.53, with 31.2 percent of participants scoring at the ceiling, and produced a slope difference of 0.001 (95 percent confidence interval, negative 0.099 to positive 0.101). Additional checks demonstrated similar consistency: incorporating attitude scores yielded a contrast of 0.027; restricting analysis to facility groups with high-confidence matches gave 0.016; limiting the cohort to facility clusters containing both extreme comparison groups yielded 0.007; and substituting work experience for age produced a slope difference of 0.010. Furthermore, an influence analysis across eighty-five leave-one-group-out refits produced primary slope differences ranging tightly between negative 0.008 and positive 0.026, showing that no single hospital cluster drove the linear findings.
Psychometric diagnostics indicated that while internal consistency appeared exceptionally high, underlying structural assumptions required cautious interpretation. Cronbach’s alpha values reached 0.973 for knowledge, 0.961 for practice, 0.970 for content-restricted practice, and 0.984 for attitudes. Despite these high alpha estimates, ordinal factor models evaluated through weighted least squares with mean- and variance-adjusted test statistics revealed inadequate absolute fit across tested single-factor specifications. Configural factor models could not establish baseline structural comparability across climate subgroups, preventing researchers from fitting formal equality-constrained measurement invariance models. A correlated two-factor model separating knowledge and practice items improved relative fit over a combined single-factor structure, but its comparative fit index of 0.792 and root mean square error of approximation of 0.131 remained well below standard psychometric acceptability benchmarks. The investigators noted that high internal consistency frequently reflects item redundancy rather than unidimensional construct validity, and shared response tendencies among participants answering dozens of agreement items could inflate observed associations between separate conceptual domains.
The authors highlighted several substantive and methodological limitations that prevent these observational findings from justifying climate-tailored training policies. Because self-reported knowledge, subjective climate ratings, and reported caring practices were captured concurrently through self-administered digital questionnaires, the cross-sectional design cannot determine causal direction, rule out social desirability bias, or untangle reverse causality. Moreover, perceived departmental caring climate was captured through a single respondent-rated item, which cannot distinguish objective organizational realities such as nurse staffing ratios, leadership styles, or institutional resource constraints. The pronounced ceiling effect in reported practice scores constrained measurable variation among supportive environments, complicating comparisons across subgroups. The authors concluded that before health systems implement climate-specific educational programs, prospective interventional trials must track independently audited patient experiences, structured behavioral observations, and longitudinal practice changes over time. Until such objective outcomes are tested across diverse clinical environments, positive survey associations between self-reported knowledge and reported practice reflect personal endorsement patterns rather than proven institutional mechanisms.
Subject of Research: Association between self-reported humanistic caring knowledge and reported practice across perceived departmental caring climates among hospital nurses.
Article Title: Self-reported humanistic caring knowledge and reported practice among nurses: variation by perceived caring climate in a multicenter cross-sectional study
Article References: Cai, Y., Li, Z., & Zhou, L. (2026). Self-reported humanistic caring knowledge and reported practice among nurses: variation by perceived caring climate in a multicenter cross-sectional study. Frontiers in Public Health, 14, Article 2003529. https://doi.org/10.3389/fpubh.2026.2003529
Image Credits: AI Generated
DOI: 10.3389/fpubh.2026.2003529
Keywords: humanistic nursing care, caring knowledge, reported caring practice, departmental caring climate, cross-sectional survey, Anhui Province, nurse workforce, cluster-robust covariance, facility fixed effects, spline modeling, fractional logit model, ceiling effects, measurement invariance, confirmatory factor analysis, nursing administration, work environment, clinical caring
Cite Scienmag News
Ophelia Keating. (November 20, 2026). Nurse Caring Knowledge Links to Reported Practice While Perceived Climate Moderation Varies. Scienmag. https://scienmag.com/nurse-caring-knowledge-links-to-reported-practice-while-perceived-climate-moderation-varies/
Ophelia Keating. "Nurse Caring Knowledge Links to Reported Practice While Perceived Climate Moderation Varies." Scienmag, 20 November 2026, https://scienmag.com/nurse-caring-knowledge-links-to-reported-practice-while-perceived-climate-moderation-varies/. Accessed 20 November 2026.
Ophelia Keating. "Nurse Caring Knowledge Links to Reported Practice While Perceived Climate Moderation Varies." Scienmag. November 20, 2026. https://scienmag.com/nurse-caring-knowledge-links-to-reported-practice-while-perceived-climate-moderation-varies/








