IRB approval

Institutional Review Board approval was obtained (Singhealth CIRB 2014/651/D) prior to the commencement of the study, which waived the requirement for individual informed consent. We retrospectively analyzed the electronic medical records of 98,685 patients aged 18 and older who underwent surgery under general or regional anesthesia between 1 January 2012 and 31 October 2016 in Singapore General Hospital, a 1700-bedded tertiary academic hospital in Singapore. These clinical records were sourced from our institution's clinical information system (Sunrise Clinical Manager (SCM), Allscripts, IL, USA) and stored in our enterprise data repository and analytics system (SingHealth-IHiS Electronic Health Intelligence System—eHINTS). Mortality data in the system was synchronized with the national electronic health records for follow-up. We excluded patients who underwent transplant and burns surgery, and evaluated only the index surgery for patients who had multiple surgeries during the study period. Surgeries ranged from minor day case surgeries to major surgeries. Surgical disciplines that were included were: cardiothoracic, orthopaedics, obstetrics and gynaecology, general surgery, otolaryngorhinology, hand surgery, neurosurgery, colorectal surgery, urology, plastic surgery, and oromaxillofacial surgery.

Our final dataset comprised of 97,443 patients. (Fig 1) Data collected include patient demographics, preoperative comorbidities indices such as the ASA-PS score[27], Revised Cardiac Risk Index (RCRI) score[28] and its components such as a history of previous cerebrovascular accidents (CVA), ischemic heart disease (IHD), congestive heart failure (CHF), diabetes mellitus (DM) on insulin; priority of surgery and surgical risk classification based on the 2014 ESC/ESA guidelines on non-cardiac surgery [29,30]. These information were routinely collected during the preoperative anesthesia assessment visit. Perioperative blood transfusion data was also obtained. Our missing data on preoperative hemoglobin is about 4.68%. Due to incomplete data, we analysed 57,808 patients (59.3% of the cohort) in the multivariate regression and 77,485 patients (79.5% of the cohort) in the cox regression.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Flowchart showing study cohort derivation.

https://doi.org/10.1371/journal.pone.0182543.g001

Procedures and definitions

Preoperative laboratory investigations were performed within 90 days before the surgery, up to the day of surgery but before the surgical start time. These include serum hemoglobin, red blood cell distribution width(RDW), mean corpuscular volume (MCV) and serum creatinine levels. Anemia was defined by the World Health Organization (WHO)’s gender-based classification of anemia severity.[31] Mild anemia was defined as hemoglobin concentration of 11–12.9g/dL in males and 11–11.9g/dL in females; moderate anemia was defined for both genders to be hemoglobin concentration between 8–10.9g/dL and severe anemia defined as hemoglobin concentration <8.0g/dL.

We defined perioperative blood transfusion as red blood cell (RBC) concentrate units given during the surgery and up to one month after the date of surgery. Preexisting chronic kidney disease was calculated based the estimated glomerular filtration rate (GFR) that was derived from the latest preoperative serum creatinine levels by the MDRD equation according to KDIGO guidelines.[32] RDW was reported as a coefficient of variation (percentage) of red blood cell volume with the normal reference range for RDW in this hospital laboratory to be 10.9% to 15.7%. Levels above 15.7% were defined a priori as high RDW. This corresponded to the 90th centile of RDW values in our study population. MCV is defined a priori as microcytic if < 80fL, normocytic if is between 80-100fL, and macrocytic if >100fL.

We followed up 88.7% of our patients to 1 year. 11.2% did not achieve 1 year follow up as their date of surgery was less than a year from the assessment of mortality rates. Nevertheless, the mean follow up duration was 258 days (± 64), with a minimum of 147 days.

Statistical analysis

Statistical analysis was done in IBM SPSS Statistics v21.0. Patient data were de-identified and analyzed anonymously. Baseline characteristics of the whole cohort was determined. Proportion of categorical variables were compared using the chi-square test, while mean values of continuous variables were compared using the one-way analysis of variance (ANOVA). Multinomial regression was performed to determine patient characteristics associated with odds of mild or moderate/severe anemia over no anemia, while binomial regression was performed to determine patient characteristics that are independently associated with higher odds of high RDW (>15.7) over normal RDW (< = 15.7). In addition, cox regression analysis was done to assess the independent effect of preoperative anemia, abnormal RDW and MCV values on 1-year mortality rates. We found the interaction term between RDW, anemia and MCV to be significant, and did further subgroup analysis of the anemia categories stratified by MCV and RDW. We confirmed the proportionate hazard assumption by plotting the individual log minus log curves for each categorical variables in the cox regression.

Anemia

The overall prevalence of preoperative anemia was 27.8% of our surgical population. The odds of moderate/severe anemia over no anemia increases with increasing age of the patient, female gender, comorbidities such as renal impairment, diabetes mellitus on insulin, congestive heart failure and higher ASA scores. Our prevalence rates are similar to that reported in other large scale studies which is between 25.3% [1], 28.7% [3] and 39.7%[2]. As we included patients undergoing minor to major, elective and emergency surgeries, our anemia prevalence differs slightly but is still within the anticipated prevalence as reported in other studies, which may have included a different assortment of surgeries in their cohort.

We found a pronounced gender difference in anemia prevalence across different age groups. Our finding corroborates well with National Health and Nutrition Examination Survey (NHANES) III study of the anemia prevalence in the civilian population in the United States[7]. In both the NHANES III study and our study, females of reproductive age (18–49 years) had a higher prevalence of anemia compared to females in the post-menopausal group (aged 50–69 years in our study, and 50-64years in the NHANES III study). This may be attributed to menstruation and childbearing. In men, the prevalence of anemia was lowest in age group 18–29 years, and rose with increasing age. Across age groups < 70 years, the prevalence of anemia was higher in women than in men. This trend reversed in patients aged 70 and above in our study, and this has also been demonstrated in other studies as well. Between ages 75 to 84 years, Salive et al[33] and Skjelbakken et al [8] estimated that 14.9% to 15.0% of men and 7.1% to 12.7% of women had WHO-defined anemia, while in the oldest age group (85 years and older), anemia was present in 29.6% to 30.7% of men and in 16.5% to 17.7% of women. Strikingly, our rates of anemia in each corresponding age and gender group seem to be almost double or triple that found in the NHANES III study. This could either be due to a different racial composition, or because our patients are pre-operative, and may have surgical conditions predisposing them to anemia.[9].

We also found that the morphology of anemia changes with age. Normocytic anemia was the predominant form of anemia in all age groups, but its proportion increased with age. The youngest age group (18–29 years) had the highest proportion of microcytic anemia compared to the older age groups, and this proportion decreased with increasing age, while macrocytosis increased with age. This suggests that the causes of anemia in the elderly and young adults are different. In the NHANES III study, one third of the older population above 65 years had anemia attributed to nutritional deficiencies, the remainder was attributed to anemia of chronic illnesses and unexplained anemia[7]. This survey also revealed that young women have higher prevalence of iron-deficiency compared to women above 50 years of age and men of the same age, which may account for the higher incidence of microcytic anemia in the female reproductive age group [34].

The population of Singapore is made up of 74.3% Chinese, 13.4% Malay, 9.1% Indians[35]. In our study, taking the Chinese race to be the reference group as it is the predominant racial group, Malays have aOR for moderate/severe anemia of 1.34 (p<0.001) while Indians have an aOR of 1.19 (p = 0.001). Racial differences in anemia prevalence is also found in the NHANES III survey, where blacks and hispanics tend to have more anemia compared to whites above 65 years of age.[7] Diet and certain inheritable genetic conditions may contribute to some of the observed racial difference in anemia prevalence in our study. Many Indians are vegetarians, and the iron [36] and Vitamin B12[37] bioavailability of a vegan diet is poor. A study of young Singaporean men registering for National Service in 1990 found that iron deficiency was the most common cause of anemia among Indians, while in Malays and Chinese, the most common cause of anemia was hemoglobinopathy, of which thalassemia was the most common[38]. In Singapore babies, alpha-thalassemia mutations was most common in Chinese (6.4%), followed by Indians (5.2%) then Malays (4.8%), while beta-thalassemia mutations were most common in Malays (6.3%), followed by Chinese (2.7%) and 0.7% in Indians[39]. With the introduction of widespread antenatal screening for thalassemia in Singapore, the incidence of babies born with thalassemia major has declined dramatically. In 2001, no babies were born with thalassemia major[40]. It is unlikely that thalassemia alone would account for the differences in prevalence of moderate/severe anemia between Malay and other races, because most of the thalassemic patients in Singapore are carriers(25.79%), have thalassemia minor (37.9%) or traits (7.43%).[40] However those with thalassemia minor or traits may have borderline hemoglobin levels which may dip further when other conditions associated with anemia sets in.

Red cell distribution width and mean corpuscular volume

Red cell distribution width reflects the degree of anisocytosis in a patient’s circulating red blood cells. We found that the presence of anemia and abnormal mean cell volume, especially microcytosis, is strongly associated (aOR 17.98, p<0.001) with elevated RDW value (>15.7%). Surveys of the civilian population in the United States also found that mean MCV decreased from the 1st to the 4th RDW quartile [15,41]. In literature, the commonest causes of microcytosis with high RDW include iron deficiency anemia and thalassemia.[42–44] In our study, macrocytosis was also weakly associated with elevated RDW (aOR 1.82, p<0.001). Common causes of macrocytosis in literature that are associated with elevated RDW include myelodysplastic syndromes, vitamin B12 and folate deficiencies [45].

In our study, the odds of elevated RDW was higher with increasing ASA score, which suggests that patients with more severe comorbidities and consequently poorer health have are more likely to have high RDW values. However of the three comorbidities that we explored, only congestive heart failure, and not renal impairment and insulin-dependent diabetes mellitus, had independent odds of high RDW (>15.7%). Numerous studies have found elevated RDW in patients with heart failure to be associated with poorer prognosis[46–49]. Other cohort studies such as the NHANES III study in the United States [15,41], and in Taiwan[16] also found a trend of higher RDW in females and with increasing age. While patients of female gender and older age (> = 70 years) did have higher mean RDW levels (Table 1) in our study, these factors were not independent risk factors for the high RDW cutoff of >15.7% (90th centile) in our multivariate analysis.

Impact on mortality

In our study, patients with mild anemia had an aHR of 1.98 (CI 1.77–2.21) for 1 year mortality while patients with moderate/severe anemia had an aHR of 2.86 (2.56–3.20). These values are adjusted for the effects of common influences such as perioperative blood transfusion, ASA score, age, surgical risk, emergency status and renal impairment. In addition to mortality, other studies have found that preoperative anemia also increases risks of adverse outcomes such as stroke and acute kidney injury in cardiac surgery[50] and length of hospital stay, intensive care admission and post-operative morbidities in the non-cardiac surgical population[1,3].

Preoperative anemia is a known risk factor for perioperative transfusion, which also has an independent adverse effect on postoperative outcomes [10]. In our analysis, we also corrected for the effect of transfusion thus lending strength to the assertion that the detrimental effects of anemia are independent of the effects of transfusion.

Elevated RDW is independently associated with an aHR of 2.34 (95% CI 2.12–2.58) for one year mortality in our study. This is consistent with findings in other studies, elevated RDW levels have been shown to be an independent marker for mortality after colorectal cancer surgery, coronary bypass surgery and in elderly patients who underwent fixation of hip fracture.[17–20,51] While we do not have data in our study on the cause of mortality, elevated RDW have been associated with an increased risk of cardiovascular events and all-cause mortality in patients with and without cardiovascular disease[14,16,41,52]. Elevated RDW is also associated with stroke occurrence[53] and increased all-cause mortality in patients admitted to intensive care[54–57].

From Fig 5, it is clear that the etiology of preoperative anemia based on different red cell morphology has a significant bearing on one year mortality. For the same degree of anemia and type of MCV, patients with high RDW had higher adjusted hazard ratios (aHR) of mortality than those with normal RDW. This is most strikingly shown in patients with no anemia and normocytosis, as those with high RDW have an aHR of 4.43 (p<0.001) for mortality compared to those with normal RDW levels. Also in our study, macrocytosis was found to be independently associated with increased aHR of 1.47 (p<0.001) for mortality, and in Fig 5, across all three groups of anemia, with normal or high RDW levels, patients with macrocytosis had higher aHR of mortality than those with normocytosis of microcytosis. Macrocytosis is uncommon in our study population—only 1.5% of our patients had macrocytosis, and its prevalence increases to 4.4–4.6% in the oldest age group. The causes of macrocytosis depends on the population surveyed, but common etiologies include alcoholism, vitamin B12 and folate deficiency, drug-induced, hypothyroidism, liver disease, myelodysplastic syndrome and aplastic anemia[45]; however we did not collect data in our patients that may explain their anemia.

These findings emphasize the incremental value of considering RDW together with MCV and hemoglobin levels when estimating the risks of mortality in a preoperative patient as the resulting risk estimates could be very different. Our findings are congruous with a cohort study of 36,292 elderly patients which found worse survival in patients with elevated RDW levels and macrocytosis, followed by normocytosis then microcytosis, in both anemic and non-anemic patients[58].

One attractive explanation for the some of the cases of abnormal MCV values and high RDW levels could be nutritional deficiency, as iron deficiency anemia and thalassemia are typically associated with microcytosis and high RDW [42–44], while myelodysplastic syndromes, vitamin B12 and folate deficiencies are associated with macrocytosis and elevated RDW[45] Despite these association, the link between nutritional deficiencies, RDW and mortality is not well supported in studies that do examine them. For instance in the InCHIANTI Study, NHANES III, and WHAS I, although a pooled meta-analysis of these studies found an association between RDW and mortality, this was not dramatically different in patients with and without nutritional deficiencies[59].

Inflammation has also been explored as an explanation for the association between high RDW levels and adverse outcomes, by reducing RBC survival and disrupting erythropoiesis through the effects of pro-inflammatory cytokines, leading to a more mixed population of RBC volumes in the circulation. Some studies have found a positive correlation between RDW and CRP and ESR levels[15,24].

Strengths

Our study is the first to examine in detail the relationship between preoperative anemia, RDW and MCV levels and how they correlate with postoperative mortality across a broad spectrum of surgeries, from low-risk to high-risk surgeries, cardiac and non-cardiac surgery, and across a broad range of patient profiles, ranging from healthy young patients to older patients with multiple comorbidities. Another strength of our study was that its duration only spanned 4 years, so this reduces biases from changes in health care practices over time which may affect mortality rates.

Limitations

As this is a retrospective cohort study, our study is not able to prove a causal relationship between RDW, anemia and mortality. Furthermore, we encountered missing data due to incompleteness of preoperative investigations, as not all patients who underwent surgery in our institution required full workup preoperatively, or may have done these investigations in a different institution that is not accessible to us. Consequently, we were only able to analyze 57,808 patients (59.3% of the cohort) in the multivariate regression and 77,485 patients (79.5% of the cohort) in the cox regression. We were fortunate that as our patient’s medical record database was synced with the National Death Registry, we were able to achieve 100% follow up with all our patients.

Another significant limitation of our study is the absence of information on the patient’s nutritional status, concomitant hematological and other non-hematological conditions that may explain the anemia, MCV and RDW levels. These information would have helped elucidate the causality between anemia, MCV, RDW and mortality. Nevertheless, the primary aim of this paper was to assess the preoperative prevalence and severity of anemia, to make screening efforts more targeted and cost effective.

Finally, our data is obtained by following up on a large cohort from a single center, which may limit the generalizability of the data to other populations locally and internationally. Nevertheless, we hope that given the large number of patients included in the study, the conclusion is robust and will encourage other population level studies to validate our results.

Future direction

Our study contributes to the growing body of literature on the association between elevated RDW levels and mortality. While this association is clear, no standardized cutoff exists to define high RDW levels. We chose a higher level of 15.7%, which corresponds to the 90th centile in our study population, and also the cutoff level for normal limit in our laboratory. Our study also illustrates that high RDW with macrocytosis and normocytosis, seem to be associated with higher risk for mortality compared to microcytosis, for the same degree of anemia. Future prospective studies can be done to determine the etiology of anemia in preoperative patients and their association with RDW and MCV levels, by evaluating the patient’s nutritional status and other known etiologies, to assess if any reversible etiologies may be present, managed and reversed.