Our study implanted a NH risk prediction model into the system, which can assist nurses to quickly identify neonates at high risk of hypoglycemia. Using the NH risk prediction model as a grading tool and CCC as the standardized language, an evidence-based and expert revision method was applied to construct an NH graded preventive nursing knowledge base, which can achieve NH risk management, and this information is the basis for assisting nursing personnel in making clinical decisions. Computer decision support technology was integrated to form an NH preventive nursing decision support system, which provides scientific decision support for health care workers through reminder, decision-making, warning, and data storage functions; reduces the incidence of NH; promotes the standardization of NH management; and realizes the risk management of NH.
Assessing the risk of a disease is one of the greatest challenges in medical sciences. Most clinical decisions are made based on physicians' personal understanding and experience; however, their expertise may not be adequate for assessing the risk of all diseases or disorders. Therefore, the risk assessment of diseases has been the focus of many research studies in recent years [25]. Maryam Ahmadi et al. [26] used an artificial neural network to construct a clinical decision support system for predicting quality of life among elderly people, and the results showed that the clinical decision support system, which was designed based on the CFBP, is an effective tool for improving the quality of life of elderly people. Nasrin Chekin et al. [27] constructed a clinical decision support system for assessing the risk of cervical cancer, and the study confirmed that the system can facilitate the process of identifying people who are at risk of developing cervical cancer. In addition, the system can help to increase the quality of health care and reduce the costs associated with the treatment of cervical cancer. Current decision support systems for hypoglycaemia risk assessment are mostly focused on adults and those with gestational diabetes. Spat et al.[21] developed an interactive clinical decision support system for assessing the hypoglycaemic response in type 2 diabetic patients. Patients used the clinical decision support system to calculate the insulin dosage to avoid manual insulin dose calculation errors. Health care professionals can use the system to determine a patient's risk of hypoglycaemia due to the use of glucose-lowering medications resulting in the risk of hypoglycaemic reactions. In 2018, Abejirinde et al. [28] developed a clinical decision support system for risk identification, including GDM risk, to perform early prediction of women at risk for GDM, use colour signals to visualize cues for risk categories with urgency of referral, and push recommendations for counselling and treatment decision-making to health care providers. In our study, a precompleted NH risk prediction model was implemented to assist health care professionals in classifying neonates who did not experience hypoglycaemia into hypoglycaemic risk classes, which allowed for the rapid identification of neonates at high risk of NH so that early preventive care measures could be taken. Ultimately, the incidence of neonatal hypoglycaemia was reduced.
As early as 2011, the Fetal and Neonatal Committee [15] suggested that targeted preventive management should be developed for newborns according to different risk groups. However, the diagnostic criteria and intervention thresholds for NH have been controversial both nationally and internationally [1–2]. According to the American Academy of Pediatrics (AAP) [29], NH can be diagnosed in neonates with blood glucose levels < 2.5 mmol/L in the first 24 h after birth and < 2.8 mmol/L in neonates 24 h after birth. Queensland Health (QLD) [12] recommends that a glucose level < 2.6 mmol/L should be considered NH, and a glucose level < 1.5 mmol/L should be considered to indicate severe hypoglycaemia. The Canadian Paediatric Society (CPS) [30] suggests that a blood glucose level < 2.6 mmol/L for exclusively breastfed, appropriate-for-gestational-age term infants and a blood glucose level < 3.3 mmol/L for high-risk term, preterm, and small for gestational age newborns can be considered to indicate NH. The 5th edition of Practical Neonatology in China [3] defines that the criterion for determining hypoglycaemia is a blood glucose level of < 2.2 mmol/L.
The American Academy of Breastfeeding Medicine (ABM) [31] suggests that a blood glucose level < 2.2 mmol/L in full-term healthy newborns and < 2.5 mmol/L in newborns with risk factors or clinical signs of hypoglycaemia are the thresholds for intervention. According to the Paediatric Endocrine Society (PES) [32], neonates with blood glucose levels < 2.8 mmol/L within 48 h of birth are at risk of hypoglycaemic brain injury, and therefore, it is recommended that blood glucose levels < 2.8 mmol/L be the threshold for intervention. Dixon KC et al. [33] showed that 88% of 135 UK national health care departments used a blood glucose level < 2.6 mmol/L as the threshold for clinical intervention. A previous cross-sectional survey conducted by our research team [34] revealed that the range of NH intervention thresholds in 21 hospitals in 13 cities ranged from 2.2 to 2.9 mmol/L, that the thresholds for hypoglycaemia intervention differed among different hospitals and even between neonatal and obstetrics departments in the same hospital, and that the lack of a uniform management specification for NH prevention in clinical practice caused a corresponding difference between the nursing staff in the prevention and management of NH.
The CNDSS is an overarching framework for the nursing process, based on nursing research, with predefined nursing diagnoses, correctly determined links between them, and patient outcome-oriented evidence-based nursing interventions that provide nurses with decision-making guidance [16]. A typical CNDSS contains 3 parts: a reasoning machine, a knowledge base, and a human‒machine interface, of which the knowledge base is the key of the whole system [35]. The knowledge base constructed in this study was based on the best evidence and used to develop the first draft of the NH-graded preventive care knowledge base, which guarantees the scientific validity of the knowledge base. In the Delphi method session, experts in the field of neonatal care were invited to evaluate the wording, expression, completeness, and usefulness of the content of the first draft of the knowledge base. When a newborn is admitted to the department, a pop-up window reminds the nurse to conduct an assessment, and the nurse only needs to check the yes or no box for the six risk factors. The system can automatically calculate whether the newborn is at high risk for hypoglycaemia. In the case of a high-risk baby, the system automatically jumps to the care plan module, where nurses can select personalized measures according to the newborn's high-risk factors, providing decision-making support and helping to promote the standardization of NH management.
Nurses' acceptance or satisfaction with the CNDSS is a key factor in applying the system [36]. Nurses can obtain patient information faster and make reasonable analyses and scientific judgment with the assistance of the CNDSS, thus reducing nursing errors and improving the quality and efficiency of clinical care. However, in the process of actual clinical application, the design of system functions, the quality of the knowledge base, and the usability, speed and flexibility of the system affects the use and promotion of the CNDSS [37, 38, 39]. For example, if the CNDSS only assesses the current risk factors for a patient and does not further provide appropriate nursing care for the risk factors, nurses will believe that this system only identifies the current problems of the patient without proposing effective solutions, which will lead to their unwillingness to continue to use the system [40]. In this study, nurses' satisfaction with the NH preventive care decision support system, which was nurse-led, involved participation throughout the process, and comprehensively promoted, was high, and the trial was conducted before the opinions and suggestions of clinical users were solicited. Nurses can efficiently complete risk screening for NH after neonatal admission, provide personalized preventive care measures according to different risk levels, and form structured nursing records, which reduces the clinical nursing workload. The NH assessment rate in our study was more than 90%, indicating that nurses have a certain degree of acceptance, recognition, and adherence to the NH preventive nursing decision support system.
Research Limitations
As the system is still in the initial stage of application, nurses are still required to manually check the risk factors, and the warning reminders for doctors are not yet perfect. We will continue to determine the application needs and collect feedback from clinical nurses in the process of using the system and improve the system in a timely manner.