Leading Comorbidity associated with 30-day post-anesthetic mortality in geriatric surgical patients in Taiwan: a retrospective study from the health insurance data

Background Elderly patients with aged physical status and increased underlying disease suffered from more postoperative complication and mortality. We design this retrospective cohort study to investigate the relationship between existing comorbidity of elder patients and 30 day post-anesthetic mortality by using International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM) from Health Insurance Database. Methods Patients aged above 65 years old who received anesthesia between 2000 and 2010 were included from 1 million Longitudinal Health Insurance Database in (LHID) 2005 in Taiwan. We use age, sex, type of surgery to calculate propensity score and match death group and survival one with 1:4 ratio (death: survival = 1401: 5823). Multivariate logistic model with stepwise variable selection was employed to investigate the factors affecting death 30 days after anesthesia. Results Thirty seven comorbidities can independently predict the post-anesthetic mortality. In our study, the leading comorbidities predict post-anesthetic mortality is chronic renal disease (OR = 2.806), acute myocardial infarction (OR = 4.58), and intracranial hemorrhage (OR = 3.758). Conclusions In this study, we present the leading comorbidity contributing to the postoperative mortality in elderly patients in Taiwan from National Health Insurance Database. Chronic renal failure is the leading contributing comorbidity of 30 days mortality after anesthesia in Taiwan which can be explained by the great number of hemodialysis and prolong life span under National Taiwan Health Insurance. Large scale database can offer enormous information which can help to improve quality of medical care. Electronic supplementary material The online version of this article (10.1186/s12877-017-0629-7) contains supplementary material, which is available to authorized users.


Background
Increased life expectancy, improvement of anesthesia safety and less invasive surgical techniques have made greater number of geriatric patients receive surgical intervention. With aged physical status and increased underlying disease, the risk of anesthesia and postoperative complication and mortality is much higher than other populations [1,2].
The main four factors of surgical risk and outcome in patients older than 65 years old are age,physiologic status,coexisting disease, and type of procedure [3,4]. Earlier studies suggest that anesthetic complications are related to age and some studies also have corroborated an association of mortality and morbidity with American Society of Anesthesiologists physical status (ASA-PS) scores. The surgical procedure itself significantly influence postoperative risk and it can be classified to low, intermediate, and high-risk surgery [5].
The ASA-PS classification introduced to clinical practice since 1941 was used worldwide to quantify the amount of physiological reserve that a patient possesses when assessed before a surgical procedure. This classification is validated as a reliable independent predictor of medical complications and mortality following surgery in peer review articles [6,7]. However, the ASA-PS scale has unreliability due to its inherent subjectivity which resulted in different ASA class rated in one patient by different anesthesiologists [8]. It is useful but lack of scientific precision.
To date, national health insurance database in Taiwan has recruited most patients' information and medical record for more than 10 years. Several studies have been published by using the reimbursement claims data of Taiwan's national Health Insurance [9][10][11]. We design this retrospective cohort study to investigate the relationship of existing comorbidity of geriatric patients who came for anesthesia with 30 day post-anesthetic mortality rate by using International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM). We hope to investigate the impact of different underlying comorbidity of the geriatric patient on post-anesthesia mortality.

Data base
Taiwan's National Health Insurance was put into practice since 1995 and covered more than 22.6 million residents in Taiwan. Taiwan's National Health Research Institutes established a National Health Research Database which record all in-patient and out-patient medical services for research [9]. This study used the 1 million Longitudinal Health Insurance Database in 2005 (LHID), which means 1 million patients were randomly enrolled in 2005 and the longitudinal database included all the issue from 2000 to 2010. The database was decoded with patient identifications to protect patients' privacy and scrambled for further public access. This study was approved by National Taiwan University Hospital Ethics Committee (201411078RINC) and inform consent was waived.

Study sample
The study sample is the patients aged above 65 years old and received anesthesia between 2000 and 2010. There were 420,848 index surgery requiring anesthesia in this period, including general anesthesia 304,308 times, brachial plexus block 5518 times, spinal anesthesia 85,888 times, and epidural anesthesia 2,5134 times. We defined mortality as death date appeared within 30 days after index surgery whether in hospital or not. There were 2324 death and 418,524 survival after index surgery. Due to tremendous difference in population, we use age, sex, type of surgery to calculate propensity score [12,13] and match death and survival group with 1:4. Among them, there were 6729 patients aged above 65 years old and 1401 patients were dead (Fig. 1).

Key variable of interest
We use International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM) appeared 2 years before index surgery in our database as comorbidity. The definition of comorbidity means the patient was diagnosed for more than 3 times and the interval was more than 28 days which including ischemic heart disease, hypertension, heart failure, vascular disease, respiratory disease, disease of liver and biliary tract, disease of GI system, urinary disease, endocrine disease, musculoskeletal disease, infectious disease, CVA or trauma, cancer, other disease..(Additional file 1) Due to disease categorization is complex, we therefore aggregated codes into disease group to resemble clinical preanesthetic usage. This process was conducted independently by three anesthesiologists.

Statistical analysis
The difference of comorbidity in death and survival group 30 days after index surgery was analyzed by Chi-Square test. We use conditional logistic regression to correct age, gender, type of surgery and other comorbidity, then analysis the correlation of comorbidity with death. Multivariate logistic model with stepwise variable selection [14] was employed to investigate the factors affecting death 30 days after anesthesia. We perform calculation by SAS statistical package (SAS System for Windows, Version 9.3; SAS Institute Inc., Cary, NC).

Results
More than one hundred codes were given out when we count all the comorbidity ICD-9 code in death group. Seventy three codes were selected after aggregation by expertise. (Additional file 1) All the comorbidity was compared by chi square test under 1:4 ratio by matching age, sex, type of surgery as Table 1 listed. Age and sex were both statistically significant after propensity score matching. The crude odds ratio and adjusted odds of each comorbidity ( Table 2) was counted and then 37 leading comorbidities (Table 3) which can independently predict 30 days post-anesthetic mortality in geriatric patients were ranked by multivariate logistic model with stepwise variable selection. In our study, the leading comorbidities predict post-anesthetic mortality is chronic renal disease, acute myocardial infarction, and intracerebral hemorrhage.

Discussion
With better medical quality and living condition, geriatric patient population is growing and often pose a significant challenge in surgery and anesthesia. Geriatric patients are relative fragile and also develop more complication after anesthesia than general population [1,15]. The most common postoperative complication is pulmonary complication and the secondary is cardiac event, leading to longer hospitalization and increased mortality. In previous study in Taiwan, relationship between postoperative complications and mortality risk was established, but there was no analysis between preoperative comorbidities and post-operative mortality. The leading preoperative comorbidities were listed as following: Hypertension, Diabetes mellitus, Coronary artery disease, Pulmonary disease, Malignancy, Hepatic dysfunction, and Renal dysfunction. Detailed evaluation and better communicating the aforementioned risk factors to these patients before operation are suggested for improving anesthesia quality and surgical outcomes [16].
A comprehensive geriatric assessment including Activities of Daily Living (IADL), cognitive function, nutrition status, and past medical history were used to predict postoperative morbidity and mortality in geriatric patients who received elective surgery [17,18]. They came to a conclusion that aging itself not increase surgical risk, rather, the increasing prevalence of chronic disease and the deterioration of the organ's functions, might increase the risk of postoperative mortality. Geriatric patients tend to carry more than one comorbidity and it is a risk factor for functional decline, disability, dependency, and institutionalization. Risk of functional decline and deterioration of the organ's functions following comorbidities rather than age itself play an more important role in geriatric patients surgical risk assessment.
In 2015, several large scale study concerning postoperative morbidity and mortality were published, including using multidimensional frailty score to predict postoperative complications in older female cancer patients [18], peer review reporting ASA classification as a reliable independent predictor of medical complications and mortality following surgery [7], a retrospective cohort study using national anesthesia clinical outcome registry [19] on perioperative mortality in 2010 to 2014, the effect of adding functional classification to ASA status for predicting 30-day mortality [20], and newly established preoperative score to predict postoperative mortality (POSPOM) [21]. All the above indicate that the lacking and desiring of an objective preoperative evaluation tool to predict perioperative risk and morbidity.  This is the first retrospective cohort study investigating relationship of comorbidity of elder patients with 30 day post-anesthetic mortality rate using International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM) from Taiwan Health Insurance database. We solely investigated disease code in our study to diminish other man-made bias in the health insurance database and aggregated them into 73 comorbidities by expertise to include most comorbidities. We also adopted death date to include both in-hospital and out-of-hospital death to avoid mortality bias. We used 1:4 propensity score matching case control to select comparable controls, but there were still significant differences in age and sex proportions (p < 0.001). A possible explanation is that the large sample size in the present study might be the reason for the statistical significance, but not clinically significant [22]. For example, the difference between 76 years old in non-death group and 78 years old in death group. Multivariate logistic model with stepwise variable selection was then applied to analysis the ability of comorbidities to predict postoperative mortality. From the 33 comorbidities, the leading comorbidity predicts post-anesthetic mortality in order is chronic renal failure, acute myocardial infarction, and intracerebral hemorrhage.
In the past, cardiovascular disease was regarded as the leading comorbidity that contribute to aged patients' functional decline [23]. Due to poor cardiopulmonary reserve, limited daily activity and function capacity resulted in disability and institutionalization. However, chronic renal dysfunction was found to have better predicting ability to postoperative mortality than myocardia infarction by stepwise variable selection in our study. This can be explained by the increasing number of hemodialyzed patients in Taiwan after National Health Insurance put into practice. Due to low cost of insurance fee, patients with chronic renal failure received more medical care and have longer life span. However, multiple organ system deteriorated rapidly and thromboembolic events increased with longer duration of hemodialysis [24]. Amputation and artificial vascular surgery put these patients in a higher mortality rate after anesthesia [25]. Chronic kidney disease associated with increased risk of death, increased cardiovascular events and hospitalization was proven [26] and it also increased adverse outcome after elective orthopedic, general, and vascular surgery [27].
The secondary leading comorbidity predicting postanesthetic mortality was acute myocardial infarction compatible as other studies. Risks related to the patient and related to surgery are both high for unstable hemodynamic status and emergent coronary artery bypass. A recent myocardial infarction remains a significant risk factor for postoperative MI and mortality and postponing elective operation after optimizing medical treatment is suggested [28]. Intracerebral hemorrhage was the tertiary leading comorbidity which is correlated with hemorrhagic stroke and traumatic injury accompany with poor outcome. Intracerebral hemorrhage is the most devastating type of stroke leading to greatest mortality and it is also an important public health problem leading to high rates of disability in geriatric patients [29]. Post-operative mortality is high in patients diagnosed as intracerebral hemorrhage undergoing blood evacuation.
In Current era of informative age, large scale of medical data was stored and established as a database in the national health insurance institute. From that, enormous amount of information can be acquired and work up. The limitation of our study is that our database is 1 million Longitudinal Health Insurance Database in 2005. The population is small and the data is old. The international classification of disease(ICD-9) had revised to 10th version and aggregation of ICD-9 codes made man-made bias. Besides, functional classification of ASA and geriatric dysfunction assessment were not included in the database of National Taiwan Health Insurance. Better registration system and further studies were warranted.

Conclusions
We design this study to present the leading comorbidity contributing to the postoperative mortality in