The approval for this retrospective study (no. 2017-013) was provided by the Ethical Committee of Kumamoto Central Hospital, Kumamoto, Japan, on December 27, 2017. Patients who underwent osteosynthesis or hip hemiarthroplasty for a hip fracture under general or spinal anesthesia performed between January 1, 2012, and December 31, 2018, at Kumamoto Central Hospital were eligible for this study. All data were obtained from the patients’ medical records without personal information. Informed consent from the patients was therefore waived, based on the Ethical Guidelines for Epidemiological Studies issued jointly by the Ministry of Health, Labour and Welfare and the Ministry of Education, Culture, Sports, Science, and Technology of Japan.
Data were collected based on the variables required for the GNRI and the CONUT. The GNRI requires serum albumin level and body weight, and the CONUT requires serum albumin level, total cholesterol level, and total lymphocyte count. A patient’s body weight was measured on the bed with weight meter automatically at his or her admission day. The GNRI was calculated as follows : GNRI = 14.89 × serum albumin level (g/dl) + 41.7 × present body weight/ideal body weight. The ideal body weight was defined as (height (m))2 × 22. GNRI was classified as severe (< 82), moderate (≥ 82, < 92), low (≥ 92, < 99), and normal (≥ 99) . The CONUT was a sum of the scores based on serum albumin (0, 2, 4, 6), total cholesterol level, and total lymphocyte count (0, 1, 2, 3, for each) and classified as severe (≥ 9), moderate (5–8), low (2–4), and normal (0–1) .
Factors presumably associated with postoperative complications including patient characteristics were also extracted from the medical charts: patient age, gender, underlying co-morbidities, the ejection fraction measured by echocardiography, the presence or absence of dementia, the waiting period prior to surgery, the method of anesthesia, the precise type of surgery, the operation time, and the anesthesia time. At our hospital, the method of anesthesia is left to the individual anesthesiologist. When the patient’s preoperative condition was stable, general anesthesia was the first choice. Conversely, when the patient had one or more serious underlying co-morbidities such as chronic heart failure or pulmonary disease, spinal anesthesia was administered.
In this study, we set the 180-day mortality as the endpoint, and we collected patients’ biochemical data which was sampled at their hospital admission day. When a patient dies after leaving our hospital, the family doctor, the nursing home staff, or the family always informs us the information. And that is certainly recorded in the patient’s medical chart by our hospital staff. We divided the patients into two groups: those who were still alive at 180 days post-surgery (the survivor group) and those who had not survived as of 180 days post-surgery (the non-survivor group), and we analyzed the patients’ characteristics, the GNRI value, and the CONUT value between the two groups. We also divided the patients into the four groups (severe, moderate, low, and normal) according to the grade of malnutrition using a previously established threshold [6, 7], and we compared the 180-day mortality among the four groups. We assessed the power of the GNRI and CONUT values to distinguish patients who died ≤ 180 days post-surgery from those who did not, by calculating the area under the receiver operating characteristic curve (AUC).
All statistical analyses were carried out using the software program Excel Tokei 2012 (Social Survey Research Information, Tokyo). Intergroup differences were assessed with the χ2 test with Yates’ correlation for continuity in category variables. The Mann-Whitney U test was used to test for differences in continuous variables. Differences of p < 0.05 were considered significant. Descriptive data are presented as the mean ± standard deviation (range).
We assessed the power of a model to distinguish patients who died within 180 days after the hip fracture surgery from those who did not by calculating the area under the receiver operating characteristic curve (AUC) . The AUC value ranged from 0.5 to 1.0 and the greater the AUC, the better the model. An AUC of 1.0 indicates a perfect model that has 100% sensitivity and 100% specificity. An AUC of 0.5 indicates a model that is completely ineffective in differentiating between real cases and non-cases. In addition, we analyzed the statistical correlation between these two models by Spearman’s rank correlation (ρ).