Enhanced Nursing Intervention, Glycemic Variability, And In-Hospital Outcomes In Type 2 Diabetes Mellitus: A Retrospective Cohort Study

Summary

This retrospective propensity score-matched study of 100 patients with type 2 diabetes mellitus found that enhanced nursing intervention reduced glycemic variability and improved short-term in-hospital outcomes compared with usual care.

Abstract

This study investigated the association between enhanced nursing intervention and glycemic variability, as well as short-term hospitalization outcomes, in patients with type 2 diabetes mellitus (T2DM). A retrospective cohort design was employed, consecutively enrolling hospitalized patients with T2DM from December 2022 to December 2025. Propensity score matching (1:1) was performed to balance age, sex, disease duration, and baseline medication use, resulting in 50 patients receiving enhanced nursing care and 50 receiving routine care. Enhanced nursing included dietary and exercise guidance, psychological counseling, self-management education, and continuous glucose monitoring. Primary outcomes included glycemic variability metrics, including standard deviation of blood glucose, mean amplitude of glycemic excursions, largest amplitude of glycemic excursions, coefficient of variation, and time in range. Secondary outcomes included hospitalization parameters, including length of stay, readmission rate, hypoglycemic events, and nursing satisfaction. After matching, baseline characteristics were comparable between groups. Compared with the routine care group, the enhanced nursing group demonstrated significantly lower glycemic variability indicators and reduced fasting and postprandial glucose levels (p < 0.05). Hospitalization outcomes also improved significantly, including shorter length of stay (p < 0.05), lower 30-day readmission rate (2.0% vs. 14.0%, p = 0.027), and fewer hypoglycemic events (6.0% vs. 24.0%, p = 0.012). Multivariate regression analysis identified enhanced nursing as an independent protective factor associated with reduced glycemic variability (B = −7.892, p < 0.001) and shorter hospitalization duration (B = −3.112, p < 0.001). These findings suggest that enhanced nursing intervention is associated with improved glycemic stability and better short-term hospitalization outcomes in patients with T2DM.

Introduction

Type 2 diabetes mellitus (T2DM) is a common metabolic disorder with a continuously increasing global prevalence^1,2,3. According to recently published data from the International Diabetes Federation, China has the highest number of adults affected by diabetes worldwide, with T2DM accounting for >90% of cases^4,5. Clinical assessment of glycemic control has traditionally relied on glycated hemoglobin as the gold standard because it reflects average blood glucose levels over the preceding 2 to 3 months^6,7. However, the widespread use of continuous glucose monitoring (CGM) technology has revealed important limitations of focusing solely on average glucose levels, as patients with similar glycated hemoglobin levels may exhibit markedly different patterns of glycemic variability^8,9. Some patients with acceptable glycated hemoglobin control continue to experience frequent and pronounced glucose fluctuations, which are considered an independent risk factor for cardiovascular complications^10,11. Previous studies have demonstrated that glycemic excursions are more likely to induce oxidative stress and endothelial dysfunction than sustained chronic hyperglycemia, providing a mechanistic basis for the association between glycemic variability and diabetic vascular complications^12,13.

In clinical practice, glycemic fluctuations are often overlooked. Routine fingertip blood glucose monitoring cannot capture the full pattern of glucose changes throughout the day, resulting in delayed detection of many hypoglycemic episodes and hyperglycemic peaks^14,15. In hospitalized patients, glycemic variability is particularly pronounced because of factors such as physiological stress, dietary changes, and medication adjustments^16,17. Clinical observations indicate that some patients experience difficulty maintaining stable glycemic control after admission, with frequent glucose fluctuations that may prolong hospitalization and increase the burden on healthcare providers. Consequently, strategies aimed at reducing glycemic variability while maintaining stable glucose control have become an important focus in diabetes management research^18,19.

Nursing intervention is an important component of comprehensive diabetes management and plays a critical role in glycemic control^20,21. Conventional nursing care primarily focuses on monitoring vital signs and implementing medical orders, whereas enhanced nursing interventions incorporate individualized health education, dietary guidance, exercise prescriptions, psychological support, and the interpretation and application of CGM data^22,23. This multidimensional nursing approach may help patients better understand patterns of glycemic change, improve self-management behaviors, and reduce both average glucose levels and glycemic variability. However, there remains insufficient evidence regarding whether enhanced nursing interventions can significantly improve glycemic variability measures and optimize hospitalization outcomes in patients with T2DM. Most existing studies have focused primarily on glycated hemoglobin, with relatively few systematically evaluating glycemic variability parameters, and even fewer applying rigorous methods such as propensity score matching (PSM) to control for confounding bias^24,25.

Glycemic variability has emerged as an independent risk factor for diabetic complications, with evidence demonstrating that repeated glycemic excursions activate oxidative stress pathways more strongly than sustained hyperglycemia, thereby accelerating endothelial dysfunction^12,13. Accordingly, major clinical guidelines now recognize glycemic variability, together with time in range (TIR) and coefficient of variation (CV), as important metrics for comprehensive glucose assessment^10,11,20,21. Nurse-led diabetes interventions have been shown to improve glycemic outcomes effectively. A 2025 scoping review of intensive care unit nurse-led glycemic protocols reported that structured nurse-led management, including continuous intravenous insulin infusion protocols, standardized monitoring intervals, and printed or computerized decision-support tools, was associated with reductions in glycemic variability and the incidence of both hyperglycemia and hypoglycemia⁷. Similarly, a systematic review of 13 randomized controlled trials demonstrated that nursing-led digital health interventions incorporating personalized coaching and regular follow-up significantly improved glycated hemoglobin, self-management behaviors, and quality of life¹². A meta-analysis of eight randomized controlled trials further reported that nurse-led diabetes self-management education reduced glycated hemoglobin by 0.92% at 4–6 months (MD = −0.92, 95% CI: −1.44 to −0.41) and by 0.54% at more than 6 months (MD = −0.54, 95% CI: −0.86 to −0.23)¹³. Despite these advances, most existing nursing intervention studies have relied on traditional glycemic metrics, such as glycated hemoglobin and fasting glucose, rather than comprehensive glycemic variability indices including mean amplitude of glycemic excursions (MAGE), CV, and TIR. In addition, relatively few studies have applied rigorous PSM to control for selection bias, and no previous study has combined detailed glycemic variability metrics with propensity score-matched nursing evaluation in hospitalized patients with T2DM. The present study addresses these gaps by systematically evaluating the association of an enhanced nursing intervention, including daily CGM-based feedback, individualized diet and exercise guidance, psychological support, structured discharge assessment, and post-discharge follow-up, with glycemic variability and short-term hospitalization outcomes using a propensity score-matched design.

This retrospective cohort study investigated the association between enhanced nursing intervention, glycemic variability, and short-term hospitalization outcomes in patients with T2DM. After PSM was performed to balance potential confounding factors, 50 patients were included in both the enhanced nursing and routine care groups. The study compared the effects of the two nursing models on glycemic variability indicators, length of hospital stay, readmission rate, hypoglycemic events, and nursing satisfaction. The novelty of this study lies in shifting the focus from average blood glucose measurements to comprehensive glycemic variability assessment while applying PSM to reduce selection bias and strengthen analytical rigor. To our knowledge, this is the first retrospective study to combine glycemic variability metrics, including MAGE, CV, and TIR, with propensity score-matched nursing evaluation in hospitalized patients with T2DM, thereby providing evidence regarding the associations among enhanced nursing intervention, glycemic stability, and short-term hospitalization outcomes.

We hypothesized that patients receiving enhanced nursing intervention would demonstrate lower glycemic variability, reflected by reduced MAGE, CV, and largest amplitude of glycemic excursions (LAGE), as well as increased TIR, together with improved short-term hospitalization outcomes, including shorter length of stay, lower 30-day readmission and hypoglycemia rates, and higher nursing satisfaction, compared with patients receiving routine care.

Protocol

This retrospective observational study was approved by the Ethics Committee of Beijing Rehabilitation Hospital Affiliated to Capital Medical University (approval number: 2026-A-022). The study was conducted in accordance with the Declaration of Helsinki and relevant Chinese regulations governing biomedical research involving human subjects. Because the study used retrospectively collected, de-identified clinical data obtained during routine medical care, the ethics committee granted a waiver of informed consent. All patient data were anonymized before analysis, and no personally identifiable information is reported in this study. The study involved no experimental intervention, alteration of standard clinical management, or additional patient contact beyond routine medical and nursing care.

All commercial products and software are identified in the Table of Materials, which includes manufacturers, catalog numbers, software versions, and technical specifications. Generic terminology is used throughout the protocol text to minimize repetitive commercial references.

Study Design and Participants

This retrospective cohort study enrolled hospitalized patients with T2DM from the endocrinology department between December 2022 and December 2025. The study workflow is summarized in Figure 1.

Figure 1: Study flow diagram and patient selection process. Flowchart illustrating patient enrollment, group allocation, propensity score matching (PSM), intervention assignment, and outcome assessment in this retrospective study of hospitalized patients with type 2 diabetes mellitus (T2DM). Between December 2022 and December 2025, eligible patients were identified from electronic medical records and grouped according to the nursing care model received during hospitalization. After 1:1 PSM, 50 patients were included in the enhanced care group and 50 patients in the routine care group. The enhanced care group received structured nursing interventions including continuous glucose monitoring (CGM)-based feedback, individualized diet and exercise guidance, psychological support, and post-discharge follow-up, whereas the routine care group received standard inpatient diabetes nursing care. Primary outcomes included glycemic variability metrics derived from CGM, including mean amplitude of glycemic excursions (MAGE), coefficient of variation (CV), and time in range (TIR). Secondary outcomes included length of hospital stay, hypoglycemic events, readmission rate, and nursing satisfaction. Please click here to view a larger version of this figure.

Electronic medical records identified 187 eligible patients. Patients were initially categorized according to the nursing intervention received during hospitalization: enhanced nursing care or routine care. PSM (1:1 ratio; caliper = 0.02) was subsequently performed to balance baseline characteristics and reduce selection bias. Following matching, 41 patients from the enhanced care group and 46 patients from the routine care group were excluded because no suitable matches were identified. The final matched cohort included 50 patients in each group.

For analyses involving CGM-derived glycemic variability metrics, additional data quality criteria were applied. Patients were required to have at least 70% valid CGM readings over 72 consecutive hours (see CGM protocol section). After quality assessment, three patients were excluded because of insufficient valid CGM data, including one patient from the enhanced care group and two patients from the routine care group. Consequently, 97 patients (49 enhanced care and 48 routine care) were included in the final glycemic variability analysis, as shown in Figure 1.

Group Assignment Criteria:

Patients were assigned to the enhanced care group only if nursing records documented both daily individualized blood glucose feedback interviews based on CGM data and structured discharge education with competency assessment. Patients who did not receive either of these intervention components were assigned to the routine care group. Cases with incomplete, partial, or unclear documentation were excluded from the analysis.

Specialized Nurse Training:

Enhanced nursing interventions were delivered by certified diabetes specialist nurses who completed a standardized institutional training program consisting of 40 hours of theoretical instruction and 80 hours of supervised clinical practicum. Training content included CGM interpretation, insulin pump management, dietary and exercise counseling, psychological support, and diabetes self-management education. Certification required successful completion of written and practical examinations, and competency was reassessed annually throughout the study period. Detailed assessment and documentation forms are provided in Supplementary Tables 1–8 and Supplementary Files 1–4.

Exclusion and Inclusion Criteria

Patients were included if they met all of the following criteria²⁶: (1) age ≥18 years; (2) diagnosis of T2DM according to the Chinese Guidelines for the Prevention and Treatment of Type 2 Diabetes; (3) diabetes duration ≥6 months based on the confirmed diagnosis date in the medical record; (4) receipt of a stable glucose-lowering regimen, including oral antidiabetic agents, insulin therapy, or both, for at least 3 months before admission; (5) hospital stay ≥5 days; and (6) availability of either CGM data for at least 72 consecutive hours or a minimum of seven fingertip blood glucose measurements per day during hospitalization to allow accurate calculation of glycemic variability measures.

Patients were excluded if they met any of the following criteria^27,28: (1) acute diabetic complications, including diabetic ketoacidosis or hyperosmolar hyperglycemic state; (2) severe hepatic or renal dysfunction, malignant tumors, or severe psychiatric disorders; (3) pregnancy or lactation; (4) absence of pharmacologic glucose-lowering therapy during hospitalization (diet-only management); or (5) incomplete clinical records preventing extraction of primary outcome variables.

Study Design and Nursing Protocol

This retrospective cohort study extracted clinical and nursing data from the hospital electronic medical record system, nursing documentation records, laboratory information system, and CGM exports. Patients were categorized according to the type of nursing care received during hospitalization: routine care or enhanced nursing care. All nursing interventions were performed by certified diabetes specialist nurses from the endocrinology department who had completed standardized institutional training and followed the hospital’s Diabetes Care Work Manual.

All diabetes-related clinical and nursing documentation was completed according to institutional documentation policies. Electronic medical records, nursing logs, medication administration records, CGM exports, and laboratory data were reviewed retrospectively. No patients were contacted directly during the study. Institutional policy required completion of diabetes-specific nursing documentation within 24 h, weekly nursing audits with a target completion rate >95%, and double-signature verification for critical procedures, including insulin administration and insulin pump setting adjustments, to ensure data accuracy and traceability.

The institutional Diabetes Care Work Manual served as the standardized operating procedure for diabetes nursing management throughout the study period. The manual included protocols for admission assessment, blood glucose monitoring, medication administration, CGM interpretation, individualized dietary and exercise counseling, insulin pump management, hypoglycemia prevention and treatment, psychological assessment using the Generalized Anxiety Disorder-7 (GAD-7) scale, discharge education, competency assessment, and post-discharge telephone follow-up. Detailed assessment forms, intervention templates, competency checklists, and quality-control records are provided in Supplementary Tables 1–8 and Supplementary Files 1–4.

Routine Care Group:

Patients in the routine care group received standard nursing management during hospitalization in the endocrinology ward. On admission, demographic information, medical history, and vital signs were recorded as part of the standard nursing assessment. Nurses introduced the ward environment and provided routine hospitalization instructions. During hospitalization, blood glucose monitoring was performed according to physician orders using a bedside blood glucose monitoring system. Fingerstick blood glucose measurements were routinely obtained before breakfast, 2 h after meals, and at bedtime^29,30.Blood glucose results were documented in the nursing record system. When glucose values were <3.9 mmol/L or >16.7 mmol/L, nurses notified the responsible physician promptly and implemented treatment according to medical orders. Oral antidiabetic medications and insulin therapy were administered according to physician instructions. Before medication administration, nurses verified patient identity, medication type, dosage, and administration route according to institutional safety procedures. Subcutaneous insulin injections were routinely administered in the abdomen or lateral thigh using insulin injection devices. Group health education sessions were conducted three times weekly (Monday, Wednesday, and Friday afternoons) and included basic diabetes education regarding diet, medication adherence, glucose monitoring, and hypoglycemia prevention. Participation was voluntary, and attendance was recorded in the nursing documentation system. Before discharge, patients received routine discharge education, including medication instructions, outpatient follow-up scheduling, and guidance regarding home blood glucose monitoring. Standardized discharge instruction leaflets were provided. No structured active follow-up after discharge was routinely implemented in the routine care group during the study period.

Enhanced Nursing Care Group

Patients in the enhanced care group received systematically enhanced nursing intervention in addition to all routine care components. Clinical and nursing data for this group were obtained from electronic nursing records and CGM system exports. The intervention framework was developed according to Bandura’s Social Cognitive Theory and the principles of Diabetes Self-Management Education and Support (DSMES). The intervention targeted self-efficacy, behavioral reinforcement, goal setting, and environmental support through structured nursing feedback and individualized education. Detailed intervention workflows and standardized documentation forms are provided in the supplementary materials and the institutional Diabetes Care Work Manual. Within 2 h of admission, a certified diabetes specialist nurse completed a comprehensive admission assessment that included disease history, medication use, self-management ability, psychological status, and social support evaluation using standardized assessment forms (Supplementary Table 1).

Continuous Glucose Monitoring:

Within 24 h of admission, patients initiated CGM using the CGM system listed in the Table of Materials. Alarm thresholds were standardized at 3.9 mmol/L for hypoglycemia and 13.9 mmol/L for hyperglycemia³¹. CGM preprocessing and missing-data handling procedures are described separately in the data processing section. Each morning, the responsible nurse reviewed and exported the previous 24 h CGM report, including mean glucose, glucose fluctuation range, TIR, and hypoglycemic events. CGM reports were incorporated into the medical record for physician review and nursing feedback discussions.

Daily Blood Glucose Feedback Interviews:

According to nursing records, patients in the enhanced care group received individualized blood glucose feedback interviews once daily between 16:00 and 17:00. Interviews were conducted either at the bedside or in a private counseling room while patients were in a clinically stable condition. To maintain intervention consistency, all interviews throughout the study period were performed by the same team of certified diabetes specialist nurses. Each interview lasted approximately 10–15 min and followed a standardized intervention guide that included: (1) review of the previous day’s CGM curve to identify glycemic excursions; (2) discussion of dietary intake, physical activity, medication adherence, and emotional stressors associated with glucose fluctuations; (3) patient self-evaluation of glycemic control; and (4) collaborative development of up to two specific action plans for the following day. A standardized documentation form (Supplementary Table 2) was completed after each interview and included patient identification number, interview date and duration, CGM summary metrics, discussion checklist items, patient-identified problems, agreed action plans, and nurse verification signature. Key intervention points were additionally documented in the nursing record sheet. For patients receiving insulin pump therapy, the bedside insulin pump information card displayed at the foot of the bed was updated daily by the responsible nurse with the current basal and premeal insulin doses and signed for verification by patients and family members.

Insulin Pump Management:

Patients requiring continuous subcutaneous insulin infusion used the insulin pump system. Basal rates and bolus doses were prescribed and adjusted exclusively by the attending endocrinologist according to standardized titration targets, including fasting glucose 5.6–7.8 mmol/L and postprandial glucose <11.1 mmol/L. Nurses performed infusion-site inspections twice daily, including assessment of puncture-site condition, local redness or swelling, fixation integrity, and tubing security. Findings were documented in the nursing record system each morning and evening. Nurses additionally verified pump programming against physician orders, documented daily pump history, and reported persistent abnormalities to the treating physician. Independent adjustment of pump settings by nurses or patients was not permitted. All pump-related modifications were documented in both the electronic medical record and the dedicated insulin pump management form (Supplementary Table 3). A laminated bedside safety instruction card (“Three Musts and Six Don’ts”; approximately 10 cm × 15 cm) was displayed in large-font format for all patients receiving insulin pump therapy.

The “Three Musts” included: (1) checking the infusion site twice daily for redness, swelling, or leakage and reporting abnormalities immediately; (2) confirming programmed basal and bolus doses with the nurse before meals; and (3) keeping the pump dry and securely fixed to avoid accidental traction on the tubing.

The “Six Don’ts” included: (1) not independently changing pump settings, including basal rate, bolus dose, or alarm thresholds; (2) not disconnecting the pump without notifying nursing staff; (3) not showering or bathing without appropriate waterproof protection; (4) not operating the pump if the screen was cracked or wet; (5) not ignoring pump alarms and notifying the nurse immediately if alarms occurred; and (6) not reusing infusion sets or extending replacement intervals beyond 48 h. The card also included the ward phone number and the nurse station extension for urgent inquiries.

Structured Psychological Support:

Nurses responsible for psychological intervention completed a standardized 16 h certified training program in psychological support and diabetes-related counseling. Psychological assessment was performed on hospitalization days 3 and 5 using the GAD-7 scale. Patients with GAD-7 scores ≥5 initiated a structured 3-day psychological intervention protocol consisting of one daily session lasting approximately 15–20 min. Each intervention session followed a standardized framework that included: (1) active listening to patient concerns; (2) psychoeducation regarding the relationship between glycemic control and emotional status; (3) identification and cognitive restructuring of negative thoughts; (4) development of one specific coping strategy for the following 24 h; and (5) referral for additional psychological evaluation if the GAD-7 score remained ≥7 or if suicidal ideation was identified. All intervention sessions were documented using a standardized psychological intervention log that recorded session duration, counseling content, coping strategies, and referral information (Supplementary File 1). Patients with GAD-7 scores <5 received routine emotional support without structured psychological intervention sessions.

Structured Dietary Intervention:

Daily caloric requirements were calculated using the Harris–Benedict equation multiplied by an activity factor (1.2 for bedrest and 1.3 for ambulatory patients). Macronutrient distribution targets included 45%–55% carbohydrate intake from low-glycemic-index sources, 15%–20% protein intake with at least 50% high-quality protein sources, and 25%–30% fat intake with saturated fat comprising <10% of total calories. Individualized dietary adjustments were made according to patient preferences, cultural dietary habits, and clinical conditions. Patients received printed personalized meal plans and food-portion photo guides. Representative meal plans and standardized dietary templates are provided in Supplementary File 2. Dietary planning followed the food exchange portion method described in the National Health Commission 2024 Nutrition and Exercise Guidelines for Hyperglycemia³². Foods were categorized into standardized exchange groups according to macronutrient composition and caloric equivalence. One exchange portion was defined as approximately 90 kcal, or for seasonings, 1 g salt or 400 mg sodium. A simplified exchange reference table for commonly used foods is provided in Supplementary Table 4. To support hypoglycemia prevention and self-management education, each patient received a bedside “Warm Heart Box” containing three glucose tablets (4 g each), two hard candies, one 200 mL orange juice box, and two crackers. A laminated instruction card describing the “15–15 rule” was included and instructed patients to consume 15 g rapid-acting carbohydrate, wait 15 min, recheck blood glucose, repeat treatment if glucose remained low, consume a snack if the next meal was more than 1 h away, and notify nursing staff. Nurses provided initial instruction regarding hypoglycemia management and replenished supplies daily as needed.

Structured Exercise Protocol:

Patients participated in the Huichun Medical Health Exercise program twice daily under the supervision of certified diabetes specialist nurses. Exercise sessions were conducted between 09:00–10:00 and 15:30–16:30, with each session lasting approximately 20–25 min. The standardized exercise protocol included three phases: (1) warm-up exercises (3–5 min), consisting of stretching and breathing exercises; (2) main exercise training (12–15 min), including arm swinging, trunk rotation, shoulder push-pull movements, leg raising, arm circling, ankle pumping, and whole-body stretching; and (3) cool-down exercises (3–5 min). Detailed step-by-step exercise instructions and movement modifications are provided in Supplementary File 3.

Exercise intensity was prescribed at a Borg Rating of Perceived Exertion (RPE) score of 11–13, corresponding to light-to-moderate intensity exercise³³. The Borg RPE scale was displayed in the exercise area, and patients were asked to report perceived exertion during supervised sessions.

Blood glucose levels were measured immediately before and after exercise sessions. Exercise was temporarily withheld when pre-exercise glucose levels were <5.6 mmol/L or >16.7 mmol/L. Patients with elevated glucose levels underwent ketone assessment before exercise participation. Bedside or seated exercise adaptations were provided for patients with limited mobility. All sessions were supervised directly by certified diabetes nurses, with an approximate nurse-to-patient supervision ratio of 1:6–8. Emergency hypoglycemia supplies and bedside glucose monitoring equipment were immediately available during all exercise sessions.

Objective criteria for temporary exercise interruption or termination were adapted from American College of Sports Medicine (ACSM) recommendations and included the following conditions: blood glucose <3.9 mmol/L or >16.7 mmol/L with positive ketones; heart rate >85% of age-predicted maximum heart rate (or >110 beats/min in patients receiving beta blockers); oxygen saturation (SpO₂) <90%; systolic blood pressure decrease ≥20 mmHg or increase >200 mmHg; diastolic blood pressure >110 mmHg; or development of concerning symptoms, including chest pain, severe dyspnea, dizziness, ataxia, nausea, leg cramps, pallor, cyanosis, or cold sweating. When exercise interruption criteria were met, the supervising nurse documented the reason for exercise cessation, vital signs, symptoms, and nursing interventions in the exercise record. If symptoms resolved within 10 min with rest alone and blood glucose stabilized between 5.6 and 13.9 mmol/L, exercise could resume at approximately half of the original intensity for up to 10 additional minutes under direct supervision. If symptoms persisted or abnormal physiologic parameters remained unresolved, the exercise session was terminated and the attending physician was notified immediately. The complete exercise workflow and movement instructions are provided in Supplementary File 3.

Standardized Pre-Discharge Competency Assessment:

Within 24 h before discharge, patients in the enhanced care group underwent a structured discharge competency assessment performed by a certified diabetes specialist nurse. The assessment evaluated insulin injection technique, blood glucose meter operation, and hypoglycemia self-management using standardized scoring criteria and predefined passing thresholds.

For patients prescribed insulin therapy, insulin injection competency was assessed using a 10-point checklist that included correct insulin type and dose verification, hand hygiene, insulin pen priming, injection-site rotation, appropriate needle insertion angle, slow injection with a 10 s hold before withdrawal, safe needle disposal, accurate dose recording, and verbalization of meal-related insulin timing. A passing score was defined as ≥8/10. Patients who failed underwent repeat bedside education followed by reassessment after approximately 30 min, with a maximum of two reassessment attempts permitted before escalation to the attending physician.

Blood glucose meter operation was assessed in all patients using a 6-point checklist evaluating preparation of testing equipment, hand washing and finger puncture technique, and successful acquisition and interpretation of glucose readings. A passing score was defined as ≥5/6. Patients who failed received repeat bedside demonstration and one reassessment within 2 h. Persistent failure prompted referral to outpatient diabetes education services.

Hypoglycemia self-management competency was also assessed in all patients using a 6-point checklist evaluating recognition of hypoglycemia symptoms, correct explanation of the “15–15 rule,” and demonstration of emergency carbohydrate use from the “Warm Heart Box.” A passing score was defined as ≥5/6. Patients who failed underwent standardized re-education lasting approximately 5–10 min followed by one repeat assessment. If competency remained inadequate, family caregivers received structured training and the patient was flagged for intensified post-discharge follow-up.

Overall discharge readiness required successful completion of all applicable assessment domains. Nurses documented individual domain scores and final discharge recommendations using the standardized Discharge Competency Checklist (Supplementary Table 5). Patients who passed signed the checklist documentation. Persistent competency deficits were documented in the electronic medical record, and community health nurses were notified to arrange follow-up assessment within 72 h after discharge.

Discharge Process:

At discharge, patients received a sealed emergency information card containing diagnosis, medication, and emergency contact information. Patients were additionally enrolled in the hospital doctor–patient communication group to facilitate post-discharge communication and continuity of care.

Post-Discharge Telephone Follow-up Protocol:

Post-discharge follow-up calls were conducted by the same certified diabetes specialist nurse who had provided inpatient nursing care in order to maintain continuity of intervention and patient familiarity. Each patient received two scheduled telephone follow-up calls: the first approximately 7 days after discharge (acceptable range: 6–8 days) and the second approximately 14 days after discharge (acceptable range: 13–15 days). Each follow-up call lasted approximately 10–15 min.

The telephone follow-up protocol followed a standardized structure that included: (1) nurse self-introduction and patient identity verification; (2) review of recent fasting and postprandial glucose values from the preceding 2–3 days, including documentation of hypoglycemic events, symptoms, timing, and corrective actions; (3) verification of medication adherence and current insulin or oral medication regimen; (4) specific inquiry regarding nocturnal or severe hypoglycemic episodes; (5) brief assessment of dietary adherence and physical activity participation; (6) individualized problem-solving guidance, including reinforcement of injection technique, snack timing, or glucose management strategies; and (7) reinforcement of follow-up appointment schedules and emergency contact procedures.

All responses were documented in real time using the standardized telephone follow-up form provided in Supplementary Table 6. Follow-up data were independently double-entered into a secure database by two research assistants, and discrepancies were resolved through third-party verification. If a patient could not be contacted after three attempts across two consecutive days, the nurse left a voicemail requesting return contact, and the unsuccessful follow-up attempt was documented in the follow-up record.

Data Verification and Discrepancy Resolution:

All data extracted from electronic medical records, nursing documentation systems, CGM exports, and laboratory information systems were independently entered into two separate spreadsheet databases by two trained research assistants. The databases were subsequently compared using a custom script developed in R statistical programming software (version 4.2.1) to identify discrepancies. Detected discrepancies were categorized as typographical errors, ambiguous source data requiring re-abstraction, or missing values. For each discrepancy, both research assistants independently re-examined the original source documents, including electronic medical records, paper nursing logs, laboratory reports, and CGM exports. If agreement could not be reached after review, a third senior investigator adjudicated the final value. All corrections and resolutions were documented in a standardized discrepancy log (Supplementary Table 7). After discrepancy resolution, a single verified analytical database was generated for all statistical analyses. Inter-rater agreement before discrepancy resolution was 96.2%, and complete agreement was achieved after final adjudication and database verification.

Laboratory Testing Protocol, Calibration, and Quality Control:

Fasting venous blood samples (5 mL) were collected from each patient between 06:00 and 07:00 on the day of admission and again on the day before discharge. Samples were processed within 2 h after collection. Laboratory analyses were performed using the automated biochemical analysis systems. Fasting plasma glucose was measured using the hexokinase method. Glycated hemoglobin (HbA1c) was measured using high-performance liquid chromatography (HPLC) with National Glycohemoglobin Standardization Program (NGSP)-standardized methodology. Lipid profiles were measured using enzymatic colorimetric methods. Liver function parameters were analyzed using International Federation of Clinical Chemistry (IFCC)-standardized enzymatic assays. Renal function testing included enzymatic creatinine measurement traceable to isotope dilution mass spectrometry (IDMS) standards and urea determination using the urease–glutamate dehydrogenase (GLDH) method. Laboratory analyzers underwent daily calibration according to institutional laboratory quality-control procedures. Two levels of commercial quality-control materials were analyzed at the beginning of each laboratory shift and after every 50 patient samples, with acceptable performance defined as results within ±2 standard deviations of target reference values. The laboratory additionally participated in the National External Quality Assessment Service (NEQAS) quality-assurance program for diabetes-related laboratory markers throughout the study period. All laboratory results were transferred automatically into the electronic medical record system after verification.

Continuous Glucose Monitoring Data Processing and Missing-Data Handling:

Raw CGM data consisted of glucose measurements recorded at 3 min intervals. CGM data quality assessment and preprocessing were performed before calculation of glycemic variability metrics. The preprocessing workflow included: (1) verification of data completeness, requiring at least 72 consecutive hours of CGM data after a 30 min sensor warm-up period; (2) exclusion of invalid readings, including values recorded during the initial 30 min warm-up period, readings flagged by the monitoring system as “signal error,” “sensor not properly inserted,” or “sensor dislodged,” and physiologically implausible glucose values <1.1 mmol/L or >33.3 mmol/L; and (3) management of missing data, in which gaps <30 min were imputed using linear interpolation, whereas gaps ≥30 min were not imputed and were excluded from analysis. For calculation of glycemic variability metrics, including MAGE, CV, LAGE, and TIR, patients were required to have at least 70% valid CGM readings within the predefined 72 h analysis window. Three patients were excluded because of insufficient valid CGM data (<70% valid readings). The final CGM analysis therefore included 97 patients (49 enhanced care and 48 routine care).

CGM Data Extraction and Processing:

CGM data were exported automatically as comma-separated values (CSV) files containing timestamped sensor glucose measurements using the software platform. Exported data files were processed using a standardized R script that screened for missing or invalid readings, performed short-gap interpolation, calculated glycemic variability metrics (MAGE, CV, LAGE, and TIR), and identified hypoglycemic events. The preprocessing workflow and analysis script are summarized in Supplementary File 4. To ensure processing accuracy, two research assistants independently processed all CGM export files. Processing discrepancies >2% between datasets were reviewed jointly and resolved by consensus with a senior investigator.

Comprehensive Admission Evaluation:

Within 2 h of admission, certified diabetes specialist nurses completed the institutional Comprehensive Evaluation Form for Diabetic Patients using standardized assessment procedures. The evaluation included: (1) demographic characteristics and diabetes history; (2) current treatment regimen; (3) self-management ability, categorized as independent, requiring assistance, or unable; (4) dietary habits; (5) physical activity level; (6) psychological assessment, including the GAD-7 scale; and (7) social support status. The self-management component generated a total score ranging from 0 to 20. The complete evaluation form and scoring system are provided in Supplementary Table 1.

Chronological Workflow from Admission to Follow-up:

The overall study workflow from admission through post-discharge follow-up is summarized in Figure 1. On the day of admission (day 1), certified diabetes specialist nurses completed the Comprehensive Evaluation Form within 2 h after hospital admission. Within the first 24 h, CGM was initiated, alarm thresholds were configured, and baseline laboratory blood samples were collected. Routine nursing care measures were implemented for all patients beginning on day 1.

For patients in the enhanced care group, individualized blood glucose feedback interviews were conducted daily from hospitalization day 2 until discharge between 16:00 and 17:00, with each session lasting approximately 10–15 min. Psychological assessment using the GAD-7 scale was performed on hospitalization days 3 and 5. Patients with GAD-7 scores ≥5 received structured psychological support sessions for three consecutive days. Individualized dietary counseling and exercise guidance were initiated by hospitalization day 3.

Supervised exercise sessions were conducted twice daily beginning on hospitalization day 2, with each session lasting approximately 20–25 min. For patients receiving insulin pump therapy, infusion-site assessment and pump-management procedures began immediately after pump initiation, including twice-daily site inspections and daily updating of bedside insulin pump information cards.

Structured discharge competency assessment was completed within 24 h before discharge. On the day of discharge, patients received emergency information cards and were enrolled in the doctor–patient communication group. Standardized post-discharge telephone follow-up calls were subsequently conducted approximately 7 and 14 days after discharge. All study variables and intervention records were collected retrospectively from electronic medical records, nursing documentation systems, laboratory systems, and CGM exports.

Handling of Protocol Deviations:

Adherence to the enhanced nursing intervention protocol was predefined as completion of at least 80% of scheduled core intervention components, including: (1) daily blood glucose feedback interviews on ≥80% of hospitalization days; (2) CGM use for at least 72 h with ≥70% valid sensor readings; (3) completion of the standardized pre-discharge competency assessment; and (4) participation in at least 50% of scheduled exercise sessions for ambulatory patients. No formal adherence threshold was defined for the routine care group. Protocol deviations in the enhanced care group, including missed interviews, premature CGM discontinuation, incomplete competency assessment, or interrupted intervention delivery, were documented by the responsible nurse using a standardized deviation log (Supplementary Table 8). Documentation included deviation type, occurrence date, reason when available, and corrective actions implemented. The overall protocol deviation rate in the enhanced care group was 7.2%. Reported deviations included incomplete CGM data collection in two patients and one missed feedback interview because of a concurrent medical procedure. No protocol deviations were identified in the routine care group.

Intervention Consistency and Quality Assurance:

The enhanced nursing intervention protocol was finalized before study initiation in December 2021 and remained unchanged throughout the study period. To maintain intervention fidelity, annual refresher training sessions, monthly protocol audits, and corrective feedback procedures were implemented throughout the study period. Standardized intervention templates, fixed intervention durations, and predefined nurse-to-patient supervision ratios were maintained consistently across all study years.

Monitoring for Temporal Confounding:

Potential temporal confounding effects were evaluated by stratifying the matched cohort according to admission year (2022–2023, 2023–2024, and 2024–2025). Baseline clinical characteristics were compared across periods using one-way analysis of variance, and no significant between-period differences were identified (all p > 0.05). Monthly protocol deviation rates remained stable throughout the study period (5.8%–8.5%). Review of hospital administrative records identified no major changes in diabetes-related clinical policies, nursing staffing, glucose-monitoring devices, or treatment protocols during the study period. In addition, no significant seasonal effects or COVID-19-related disruptions affecting diabetes care delivery were identified after 2022.

Outcome Measures

Primary Outcome Measures:

Primary outcome measures included glycemic variability parameters derived from CGM data collected over a minimum monitoring duration of 72 consecutive hours.

MAGE:

MAGE was defined as the arithmetic mean of glycemic excursions exceeding one standard deviation of the mean glucose value during the analysis period.

CV:

The CV was calculated as the ratio of glucose standard deviation to mean glucose concentration and expressed as a percentage:

LAGE:

LAGE was defined as the maximum difference between the highest and lowest glucose values recorded during the analysis period:

LAGE = Maximum glucose - Minimum glucose

TIR:

TIR was defined as the percentage of glucose measurements within the target glucose range of 3.9–10.0 mmol/L during the monitoring period:

Calculation of Glycemic Variability Metrics:

All glycemic variability metrics were derived from cleaned CGM datasets after preprocessing and missing-data management as described in the CGM data-processing section. Glycemic variability calculations were performed using statistical programming software with the iglu package (version 3.0.0) according to established definitions and published analytical standards. MAGE was calculated using the mage function with default parameters (short_ma = 5, long_ma = 31, method = "manual"). CV was calculated over the predefined 72 h analysis window. LAGE was calculated as the difference between maximum and minimum glucose values during the same monitoring interval. TIR was calculated as the percentage of valid glucose readings within the predefined target range of 3.9–10.0 mmol/L. Glycemic variability metrics were calculated only for patients with at least 70% valid CGM readings during the 72 consecutive hour analysis period.

Secondary Outcome Measures:

Length of Hospital Stay:

Length of hospital stay was defined as the total number of calendar days between hospital admission and discharge.

Diabetes-Related Readmission:

A readmission was classified as diabetes-related if the primary discharge diagnosis or principal reason for readmission documented in the electronic medical record met one or more predefined criteria based on International Classification of Diseases, Tenth Revision (ICD-10) codes, including:

1. Hypoglycemia (ICD-10: E16.0–E16.2), including drug-induced hypoglycemia;
2. Hyperglycemia (ICD-10: R73.9) or severe hyperglycemia without diabetic ketoacidosis;
3. Diabetic ketoacidosis (ICD-10: E10.10, E11.10);
4. Hyperosmolar hyperglycemic state (ICD-10: E10.11, E11.11);
5. Diabetes-related infections, including diabetic foot ulcer, cellulitis, or urinary tract infection in which diabetes was documented as a major contributing diagnosis;
6. Acute metabolic complications related to diabetes treatment, including electrolyte imbalance or uncontrolled diabetes requiring inpatient insulin adjustment.

Readmissions for elective procedures, scheduled dialysis, trauma, or clearly unrelated medical conditions were not classified as diabetes-related readmissions. All readmissions were identified using hospital administrative databases and independently verified through review of discharge summaries by two investigators. Disagreements were resolved through discussion with a third senior investigator.

Hypoglycemia Incidence:

Hypoglycemia incidence was evaluated using two complementary measures with predefined denominators. Patient-level incidence was defined as the proportion of patients experiencing at least one hypoglycemic event during hospitalization. Hypoglycemia was defined as blood glucose <3.9 mmol/L. Patient-level incidence was calculated as the number of patients with at least one hypoglycemic event divided by the total number of patients within the corresponding study group and was expressed as a percentage.

Event-rate incidence was defined as the total number of hypoglycemic events divided by the cumulative number of hospitalization days within each study group and was reported as events per 100 patient-days. Individual hypoglycemic episodes were considered distinct events if separated by at least 15 min.

For the primary between-group comparison presented in Table 6, patient-level incidence (percentage of patients experiencing at least one hypoglycemic event) was used. Event-rate incidence was additionally used to calculate the average number of hypoglycemic events per patient by dividing the total number of hypoglycemic episodes by the total number of patients in each group. Severe hypoglycemia was defined as any hypoglycemic event requiring assistance from another person for treatment, regardless of whether the patient was capable of self-treatment. All hypoglycemic events were identified from nursing documentation records and CGM logs and were independently verified by two reviewers through cross-checking of source records.

Nursing Satisfaction Assessment:

Nursing satisfaction was evaluated using the Chinese version of the Minnesota Satisfaction Questionnaire (MSQ) Short Form adapted for assessment of nursing care (MSQ Nursing Version). The questionnaire consisted of 20 items distributed across three domains: intrinsic satisfaction (12 items evaluating factors such as achievement, independence, and utilization of abilities), extrinsic satisfaction (6 items evaluating supervision, communication, and care environment), and general satisfaction (2 items evaluating overall satisfaction with nursing care). Each item was scored using a 5-point Likert scale ranging from 1 (very dissatisfied) to 5 (very satisfied). Total scores ranged from 20 to 100, with higher scores indicating greater satisfaction with nursing care services. The Chinese version of the scale has been validated previously in hospitalized patient populations. In the present study, the questionnaire demonstrated excellent internal consistency reliability, with a Cronbach’s α coefficient of 0.971 for the total scale. The complete questionnaire, including the original Chinese version and English translation, is provided in Supplementary File 4. The questionnaire was administered anonymously to all patients on the day before discharge in a private setting without the presence of clinical nursing staff to minimize response bias.

Sample Size and Statistical Power Analysis

Post hoc statistical power analysis for the primary glycemic variability outcome was performed using G*Power software (version 3.1). The analysis was based on the observed between-group difference in MAGE.

The enhanced care group demonstrated a mean glycemic fluctuation amplitude of 3.52 ± 0.98 mmol/L, whereas the routine care group demonstrated a mean value of 5.21 ± 1.45 mmol/L. The pooled standard deviation was calculated as 1.23, resulting in a Cohen’s d effect size of 1.37.

Using a two-sided significance level (α) of 0.05 and a sample size of 50 patients per group, the calculated post hoc statistical power (1 − β) was 0.996. These findings indicate that the study had sufficient statistical power to detect clinically meaningful between-group differences in glycemic variability while minimizing the probability of type II statistical error.

Statistical Analysis

Statistical analyses were performed using SPSS statistical software (version 26.0) and statistical programming software. Continuous variables were assessed for distributional normality before analysis. Normally distributed data are presented as mean ± standard deviation, whereas nonnormally distributed variables are presented as median and interquartile range. Between-group comparisons for continuous variables were performed using independent-samples t-tests for normally distributed data and Mann–Whitney U tests for nonparametric data. Categorical variables are presented as frequencies and percentages and were compared using chi-square tests or Fisher’s exact tests when expected cell frequencies were <5.

Covariate Selection and Regression Modeling:

Covariates included in multivariable regression analyses were selected a priori based on clinical plausibility and previously reported associations with glycemic variability and hospitalization outcomes. Included covariates were age, sex, diabetes duration, baseline glycated hemoglobin, and insulin use. Automated variable-selection procedures were not used. The nursing intervention group variable was forced into all regression models as the primary predictor variable.

Missing-Data Handling:

For primary and secondary outcomes, no missing data were identified among the 100 matched patients except for CGM-derived glycemic variability metrics. Three patients were excluded from CGM-based analyses because valid sensor readings accounted for <70% of the predefined 72 h monitoring period, as described in the CGM preprocessing section. For all remaining variables, including baseline covariates and secondary outcomes, the overall missing-data rate was <1%. Because of the minimal proportion of missing values, complete-case analysis was performed without data imputation. No sensitivity analysis for missing data was conducted because excluded CGM cases resulted from predefined technical quality-control criteria rather than patient-related clinical characteristics. No additional incomplete-record exclusions occurred in the final matched dataset.

Model Diagnostics:

Assumptions for linear regression analyses were assessed before model interpretation. Linearity was evaluated using partial residual plots. Independence of residuals was assessed using the Durbin–Watson statistic, with acceptable values ranging from 1.8 to 2.2. Homoscedasticity was evaluated using the Breusch–Pagan test, and residual normality was assessed using the Shapiro–Wilk test. All evaluated models satisfied predefined diagnostic criteria (p > 0.05). Multicollinearity was assessed using variance inflation factor (VIF) analysis, with all VIF values <2.5, indicating absence of significant collinearity among predictor variables. For logistic regression analysis of readmission outcomes, model fit was evaluated using the Hosmer–Lemeshow goodness-of-fit test, which demonstrated acceptable model calibration (p = 0.32). All statistical tests were two-tailed, and statistical significance was defined as p < 0.05.

PSM:

To reduce selection bias and improve baseline comparability between groups, 1:1 nearest-neighbor PSM without replacement was performed using the MatchIt package (version 4.5.0) in statistical programming software. Propensity scores were estimated using logistic regression with nursing intervention group assignment (enhanced care versus routine care) as the dependent variable. Prespecified baseline covariates included age, sex, diabetes duration, body mass index, baseline glycated hemoglobin, insulin use, sulfonylurea use, hypertension history, and hospitalization within the preceding year. Matching was performed using a caliper width equal to 0.2 times the standard deviation of the logit of the propensity score (0.02). Matching was conducted without replacement, meaning each control patient could be matched to only one enhanced-care patient. Balance after matching was assessed using standardized mean differences (SMDs), with absolute SMD values <0.1 considered indicative of negligible imbalance. After matching, all included covariates achieved SMD values <0.1, confirming adequate balance between study groups.

Results

Baseline Characteristics After PSM

After PSM, baseline characteristics were balanced and comparable between the two groups (Table 1). All patients in the routine care group received standard inpatient diabetes management, including vital sign monitoring, medication administration, blood glucose monitoring, routine health education, and discharge instructions without active post-discharge follow-up. In addition to routine care, patients in the enhanced nursing ...

Discussion

This retrospective study of 50 propensity score-matched patients with T2DM evaluated the association between enhanced nursing intervention and glycemic variability indicators. The enhanced care group demonstrated significantly improved glycemic variability parameters, shorter hospital stay, fewer hypoglycemic events, lower readmission rates, and higher nursing satisfaction compared with the routine care group. Multivariable regression analysis further identified the enhanced nursing model as an independent protective fac...

Materials

List of materials used in this article

Name	Company	Catalog Number	Comments
AiXin Health APP	Sinocare Inc., Changsha, China	Version 3.2.1	Mobile application for continuous glucose monitoring data display, alarm management, and glucose-data export RRID: N/A
Electronic medical record system	Institutional hospital information system	N/A	Used for retrieval of demographic, laboratory, medication, and nursing documentation data RRID: N/A
GAD-7 scale	American Psychiatric Association	N/A	Seven-item anxiety assessment questionnaire used for psychological screening; score range 0–21 RRID: N/A
G*Power	Heinrich-Heine-Universität Düsseldorf, Germany	Version 3.1	Statistical power and sample-size calculation software RRID: SCR_013726
Hitachi 7600-110 automatic biochemical analyzer	Hitachi, Ltd., Tokyo, Japan	7600-110	Used for fasting plasma glucose, glycated hemoglobin, lipid profile, and liver/kidney function testing; daily calibration and quality-control procedures applied RRID: N/A
iglu R package	Broll et al.	Version 3.0.0	Used for calculation of glycemic variability metrics including MAGE, CV, LAGE, and TIR RRID: N/A
Insulin pen	Novo Nordisk, Bagsværd, Denmark	NovoPen 4	Used for subcutaneous insulin administration RRID: N/A
MatchIt R package	Ho et al.	Version 4.5.0	Used for propensity score matching RRID: SCR_025618
Medtronic Paradigm insulin pump	Medtronic, Northridge, CA, USA	MMT-712EWS	Continuous subcutaneous insulin infusion system; infusion sets replaced every 48 h RRID: N/A
Minnesota Satisfaction Questionnaire (MSQ) Nursing Version	Adapted from Weiss et al. (1967)	N/A	Twenty-item nursing satisfaction questionnaire with intrinsic, extrinsic, and general satisfaction domains RRID: N/A
OK Pro blood glucose monitoring system	Xunying Optoelectronics Co., Ltd., Taiwan, China	OK-Pro-C01	Fingertip blood glucose monitoring system; electrochemical biosensor; measurement range 0.3–38.9 mmol/L RRID: N/A
R statistical software	R Foundation for Statistical Computing, Vienna, Austria	Version 4.2.1	Statistical analysis and continuous glucose monitoring data preprocessing RRID: N/A
Sinocare iCan CGM System (Second Generation)	Sinocare Inc., Changsha, China	i3	Continuous glucose monitoring system with 3 min sampling interval, factory calibration, and 15 day wear duration RRID: N/A
SPSS Statistics	IBM Inc., Armonk, NY, USA	Version 26.0	Statistical analysis software RRID: SCR_016479
Tosoh G8 HbA1c analyzer	Tosoh Bioscience, Tokyo, Japan	G8	High-performance liquid chromatography analyzer for glycated hemoglobin measurement RRID: N/A
Warm Heart Box (hypoglycemia emergency kit)	Institutional preparation	N/A	Contains glucose tablets, candies, juice, and crackers for hypoglycemia self-management RRID: N/A