- Open Access
Addition of 25-hydroxyvitamin D levels to the Deyo-Charlson Comorbidity Index improves 90-day mortality prediction in critically ill patients
© The Author(s). 2016
- Received: 21 March 2016
- Accepted: 15 June 2016
- Published: 17 June 2016
The Deyo-Charlson Comorbidity Index (DCCI) has low predictive value in the intensive care unit (ICU). Our goal was to determine whether addition of 25-hydroxyvitamin D (25OHD) levels to the DCCI improved 90-day mortality prediction in critically ill patients.
Plasma 25OHD levels, DCCI, and Acute Physiology and Chronic Health Evaluation II (APACHE II) scores were assessed within 24 h of admission in 310 ICU patients. Receiver operating characteristic curves of the prediction scores, without and with the addition of 25OHD levels, for 90-day mortality were constructed and the areas under the curve (AUC) were compared for equality.
Mean (standard deviation) plasma 25OHD levels, DCCI, and APACHE II score were 19 (SD 8) ng/mL, 4 (SD 3), and 17 (SD 9), respectively. Overall 90-day mortality was 19 %. AUC for DCCI vs. DCCI + 25OHD was 0.68 (95 % CI 0.58–0.77) vs. 0.75 (95 % CI 0.67–0.83); p < 0.001. AUC for APACHE II vs. APACHE II + 25OHD was 0.81 (95 % CI 0.73–0.88) vs. 0.82 (95 % CI 0.75–0.89); p < 0.001. There was a significant difference between the AUC for DCCI + 25OHD and APACHE II + 25OHD (p = 0.04) but not between the AUC for DCCI + 25OHD and APACHE II (p = 0.12).
In our cohort of ICU patients, the addition of 25OHD levels to the DCCI improved 90-day mortality prediction compared to the DCCI alone. Moreover, the predictive capability of DCCI + 25OHD was comparable to that of APACHE II. Future prospective studies are needed to validate our findings and to determine whether the use of DCCI + 25OHD can influence clinical decision-making.
- Vitamin D
- 25-hydroxyvitamin D
Mortality prediction scores are widely used in the intensive care unit (ICU) to predict the likelihood of survival from critical illness. These scores often assist healthcare providers in their discussions with patients and their family members about realistic expectations regarding ICU care and may be helpful in setting more patient-centered goals, assessing resource utilization, and providing higher quality of care . Mortality prediction scores are also widely used by researchers to risk adjust for severity of illness, especially in multivariable regression models [2, 3].
While several mortality prediction scores for critical illness have been developed and are in current use, the Acute Physiology and Chronic Health Evaluation II (APACHE II), first published in 1985 , is the most commonly used worldwide . However, APACHE II score calculations require measurement of specific physiological parameters within 24 h of ICU admission—this can be challenging, especially if a robust electronic medical record system is not available. Moreover, given that the score is heavily based on post-admission physiological assessments, it is not readily available at the time of admission, and it is of limited value for decision-making at the outset of critical illness.
Given the challenges of the APACHE II calculation (and for other similar mortality prediction models heavily influenced by physiologic assessments), the use of scoring systems based largely on medical history (i.e., comorbidities), despite being less accurate, are growing in popularity in the ICU literature [6, 7]. Among these, the Deyo-Charlson Comorbidity Index (DCCI) has been reported most frequently in studies of critical illness . The DCCI , although modified from the original Charlson Index , continues to have a low predictive value for mortality in ICU patient cohorts [10–14]. Various modifications of either the original Charlson Index or DCCI have been made to improve their predictive values. These modifications have typically involved the use of administrative data [15–18], and they are not specific to ICU patients. And although the inclusion of several laboratory test results in the APACHE II scoring method is integral to its superior performance over the DCCI for mortality prediction in ICU patients , the impact of adding easily measurable or readily available biomarkers to improve the predictive capabilities of the DCCI in ICU patients has been largely underexplored .
Mounting evidence suggests that vitamin D status is associated with mortality in critically ill patients [21–27]. Indeed, assessment of vitamin D status in ICU patients is becoming increasingly common in the USA and worldwide [21, 28, 29]. In general, serum 25-hydroxyvitamin D (25OHD) levels are considered the best marker of total body vitamin D status . Therefore, our goal was to determine whether the addition of 25OHD levels, measured at the outset of critical illness, to the DCCI improves mortality prediction in ICU patients.
We performed a retrospective analysis of the data from an ongoing prospective cohort study designed to assess vitamin D status in critically ill patients. A subset of these patients was previously described in studies that investigated the association of vitamin D status with duration of mechanical ventilation and 90-day mortality in ICU patients [25, 31]. For the present study, subjects were recruited from three, 18-bed ICUs (1 surgical, 1 medical, and 1 mixed surgical/medical) at the Massachusetts General Hospital (MGH), in Boston, MA. The ICUs received admissions from all surgical and medical services except for Cardiac Surgery and Cardiology, respectively. All subjects were enrolled between 06/01/2012 and 05/30/2015. MGH is a 1052-bed, teaching hospital and a level-one trauma center, which serves a diverse population in and around Eastern Massachusetts. The Partners Human Research Committee (Institutional Review Board) approved the study protocol.
Inclusion and exclusion criteria
All adult males and females, ≥18 years of age, and who were expected to require at least 48 h of critical care (as determined by the treating ICU team) were deemed eligible to participate. Informed consent was obtained either directly from subjects or appropriate healthcare surrogates. Subjects were only included in the study if blood samples to assess vitamin D status could be obtained within 24 h of admission to the ICU. Exclusion criteria included a known history of anemia at the time of ICU admission (defined as hematocrit <25 %), pregnancy or immediate post-partum status, and history of vitamin D supplementation ≥4000 IU/day. To minimize confounding from either partially treated, new-onset illness, or chronic illnesses, subjects were also excluded if they were transferred from another ICU or had been in an ICU within 1 year of the most current admission. Patients expected to transition to “comfort only measures” were also excluded.
Blood sample processing and biomarker assays
Following informed consent, fresh blood was acquired from an indwelling arterial or central venous catheter and was collected directly into an EDTA-containing tube (lavender top). The sample was immediately stored on ice and then centrifuged within 30 min to separate out plasma. All samples were centrifuged at 2300 rpm for 15 min at a temperature of 4 °C. The separated plasma was immediately transferred to polypropylene tubes and stored at −80 °C until biomarker testing was ready to be initiated. Assays were performed at the Harvard Medical School Clinical and Translational Science Award core laboratory at MGH. Plasma 25OHD (combined D2 and D3) levels were measured by enzyme-linked immunoabsorbent assay, using commercially available kits (Abbott Laboratories, Abbott Park, IL). Intra- and inter-assay coefficients of variation were both <10 %.
Clinical data collection
The MGH electronic medical records system used abstract baseline demographic information, including (1) age, (2) sex, (3) race, (4) body mass index (BMI), and (5) type of patient (surgical vs. medical). To obtain in-hospital mortality data within 90-days of ICU admission, individual electronic medical records were reviewed. For patients discharged alive from the hospital and who continued to receive care within the Partners Healthcare network (which includes MGH and its affiliates), individual electronic outpatient medical records were reviewed to document all-cause mortality within 90 days of ICU admission. All individual records were cross-referenced with the Social Security Death Index Master File to finalize 90-day mortality cases.
Descriptive statistics were tabulated for the analytic cohort. Continuous data were reported as means with standard deviations (SDs), and categorical values were expressed as proportions. To graphically represent the relationship between 25OHD levels and DCCI as well as APACHE II, we constructed locally weighted scatterplot smoothing (LOWESS) curves. LOWESS is a type of nonparametric regression, which summarizes a relationship between two variables in a fashion that initially relies on limited assumptions about the form or strength of the relationship . The rationale and methods underlying the use of LOWESS for depicting the local relationship between measurements of interest across parts of their ranges have previously been described . To investigate the association of 25OHD levels with 90-day mortality, we performed logistic regression analyses while controlling for biologically plausible covariates—as such, we developed two main models: the first controlled for (1) age, (2) sex, (3) race, (4) BMI, (5) type of patient, and (6) DCCI, while the second controlled for (1) age, (2) sex, (3) race, (4) BMI, (5) type of patient, and (6) APACHE II score.
Receiver operating characteristic (ROC) curves were constructed to assess the areas under the curve (AUC) for the ability of each predictive model to correctly identify survivors vs. non-survivors. An AUC of 0.5 suggests no predictive value, 0.7–0.8 suggests good predictive value, and >0.8 suggests excellent predictive value. Four different models were tested for their ability to predict 90-day mortality in the analytic cohort, namely (1) DCCI, (2) DCCI + 25OHD, (3) APACHE II, and (4) APACHE II + 25OHD. 25OHD levels were categorized into nationally accepted threshold levels of <10 ng/mL, 10–19.9 ng/mL, 20–29.9 ng/mL, and ≥30 ng/mL. Recent evidence suggests that these thresholds may be relevant and applicable to hospitalized and critically ill patients as well [34–36]. We assigned all plausible serial permutations of the sub-score values of the DCCI and APACHE II (0–6) to each 25OHD category in order to derive the highest AUC value. Serial number scores were assigned since the existing literature suggests that there is a near inverse linear association between 25OHD levels and adverse outcomes in hospitalized patients [34–36]. The final AUC values were compared for equality using the methodology described elsewhere . Moreover, optimal cut-points were determined from the ROC curves by identifying scores that provided the highest cumulative value for sensitivity and specificity.
In a previous study , our group demonstrated that 100 ICU patients provided sufficient power to detect a meaningful association between admission 25OHD level as well as APACHE II score and 90-day mortality. Therefore, we assumed that the current analytic cohort size (n = 310) would be adequately powered to undertake the study objectives. All analyses were performed in STATA 13.0 (StataCorp LP, College Station, TX). A two-tailed p < 0.05, and any 95 % confidence interval not spanning 1, was considered statistically significant for all analyses.
General characteristics of the study cohort (n = 310) stratified by 90-day survivors and non-survivors
90-day survivors (n = 251)
90-day non-survivors (n = 59)
67 ± 14
69 ± 18
28 ± 7
30 ± 8
Type of patient (%)
20 ± 14
12 ± 7
14 ± 7
24 ± 9
3 ± 2
5 ± 3
The optimal cutoff value for 25OHD alone was ≥3 (i.e., <10 ng/mL), with a sensitivity of 45 %, specificity of 91 %, and accuracy of 68 %, while the optimal cutoff value for DCCI alone was ≥4, with a sensitivity of 63 %, specificity of 63 %, and accuracy of 63 %. On the other hand, the optimal cutoff value for DCCI + 25OHD was ≥5, with a sensitivity of 84 %, specificity of 55 %, and accuracy of 70 %. The optimal cutoff value for APACHE II was ≥12, with a sensitivity of 95 %, specificity of 42 %, and accuracy of 69 %, while the optimal cutoff value for APACHE II + 25OHD was ≥15, with a sensitivity of 92 %, specificity of 54 %, and accuracy of 73 %.
In this retrospective, cohort study, we investigated whether the addition of 25OHD levels to the DCCI at initiation of care improved 90-day mortality prediction (vs. the DCCI alone) in ICU patients and we compared these findings to the gold standard of APACHE II assessments. We demonstrated that the addition of 25OHD levels significantly improved the AUC for the DCCI, but did not materially affect the predictive capability of the APACHE II. Moreover, our results suggest that the DCCI + 25OHD model may perform as well as the APACHE II alone for predicting 90-day mortality in ICU patients. However, given the retrospective nature of our study, the real-time utility of DCCI + 25OHD over the APACHE II requires further investigation.
Previous studies have reported AUCs for the APACHE II method, which demonstrates excellent discrimination between survivors and non-survivors of critical illness (AUC >0.80) [4, 38, 39]. AUC for the APACHE II ROC in our study is consistent with what has been reported in the existing literature. Despite its clinically acceptable predictive value, APACHE II score calculations can be cumbersome if an appropriate infrastructure is not in place. Although most commercially available electronic medical record (EMR) systems automatically calculate the APACHE II score, these products may be cost-prohibitive, especially for smaller healthcare entities . And without a robust EMR, APACHE II calculations must be performed manually by healthcare providers, which can be time consuming and therefore prone to low compliance. Furthermore, APACHE II scores may provide prognostic information too late after ICU admission. Since patients need to be admitted to the ICU for at least 24 h before a score can be calculated, the APACHE II does not help with critical decision-making for patients and families at the outset of critical illness. Indeed, many patients and families struggle with decisions to prolong medical therapy and/or undergo potentially painful as well as invasive procedures on admission to the ICU. As such, simple and timely predictive models for ICU-related mortality have the potential to assist clinicians in delivering more patient-centered, higher quality, and cost-effective care.
In 1987, Charlson et al. developed a weighted index, which took into account both the number and severity of various comorbid diseases; the purpose of the index was to develop a 10-year mortality prediction tool for the general population . In its original form, the Charlson Comorbidity Index relied on patient interviews or chart review to classify comorbid diseases. Later, Deyo et al. adapted it to accommodate the use of the International Classification of Disease, 9th Edition, Clinical Modification (ICD-9-CM) to calculate the DCCI from administrative databases . Other modifications to the original Charlson Comorbidity Index have included the use of alternate lists of specific diagnoses for comorbidities , re-assigning comorbidity weights [16, 17], and updating to the ICD-10-CM codes from ICD-9-CM codes [16, 18]. Recently, an increasing number of publications in acute care medicine have used the DCCI to risk-adjust for severity of illness [34–36, 41, 42]—indeed, mounting evidence suggests that the DCCI may have excellent in-hospital, 30-day, and 1-year mortality prediction in hospitalized patients . Regarding the ICU, DCCI alone appears to have low predictive value (which is also observed in the present study) , but in combination with administrative data that reflect acute physiological processes (e.g., admission diagnosis, need for mechanical ventilation, initiation of renal replacement therapy), the AUC of the modified DCCI may be equivalent to the APACHE II . While such adjustments are helpful for research purposes, they do not apply to real-time decision-making at the outset of critical illness. The use of non-administrative data, such as readily available or easily measured biomarkers, in conjunction with the DCCI has been reported to improve mortality prediction in dialysis-dependent renal failure patients  but until now has largely remained unexplored in the ICU.
Our study suggests that using easily obtained biomarker data, more specifically 25OHD levels at initiation of care, may significantly improve the predictive value of the DCCI for 90-day mortality in ICU patients, potentially making it a feasible alternative to the APACHE II score, especially when the necessary acute physiologic assessments are not readily available. Indeed, several studies suggest that 25OHD levels at initiation of critical care are inversely associated with mortality [25, 29, 45]. While vitamin D has traditionally been thought to play a vital role in bone , cardiac [47, 48], and muscle health , primarily by maintaining calcium balance, recent data support a more pleiotropic effect of vitamin D on general health . With regards to critical illness, vitamin D plays a major role in maintaining immune heath. Recent studies have demonstrated that cells of the innate and adaptive immune system express the vitamin D receptor. Low 25OHD levels are associated with depressed macrophage phagocytosis, attenuated chemotaxis, and proinflammatory cytokine production . Macrophages activated through the vitamin D receptor by 1,25-dihydroxyvitamin D (the most hormonally active vitamin D metabolite) upregulate expression of cathelicidin . Cathelicidins are endogenous antimicrobial peptides that are active against a broad spectrum of infectious agents, such as bacteria, viruses, fungi, and mycobacteria . Moreover, cathelicidins are highly expressed by epithelial cells at natural barrier sites (e.g., skin, lungs, gut) and may represent an important first line of defense for the innate immune system . In addition, vitamin D is important for the interferon-γ-dependent T cell responses to infection  and therefore may play an important role in preventing immunoparalysis . And finally, in animal models, vitamin D supplementation has been shown to improve coagulation variables and inhibit endotoxemia [56–58].
It is important to note that at present, it is not standard practice in most institutions worldwide to routinely obtain 25OHD levels in ICU patients. Traditionally, this has been due to a lack of evidence suggesting that assessment of vitamin D status might be helpful in critically ill patients; however, recent evidence supports the notion that 25OHD may be an important biomarker in this patient cohort [30, 35, 36]. Other major barriers to readily assessing vitamin D status in hospitalized patients worldwide have been the lack of in-house laboratory equipment, batched processing protocols, and the cost of the 25OHD assays themselves. Given the intricacies of accurately measuring steroid hormone levels, many facilities outsource such measurements, and as such, results may take several days to be finalized. Similarly, if the overall volume of testing is insufficient, many facilities process samples at dedicated time intervals—which may present a delay in obtaining results for bedside clinical decision-making. Led by the US National Institute of Standards and Technology and the National Institutes of Health, a global Vitamin D Standardization Program (VDSP) has been launched to assist facilities in ensuring the quality of their in-house assays and results [59–62]. Moreover, the recent availability of various point-of-care (POC) vitamin D testing devices has now made it feasible to obtain 25OHD levels to allow for real-time, bedside clinical decision-making [63–65]. These devices require very small amounts of blood (some are similar in design to POC glucose testing units for diabetics), are very affordable, and provide results within 10 min.
Although we present intriguing evidence regarding the addition of 25OHD and the DCCI to significantly improve mortality prediction in ICU patients, it is important to be mindful of potential limitations of our work. Retrospective cohort studies such as this cannot establish causation, but it can highlight the existence or absence of associations and thereby direct future research. Furthermore, observational studies may be limited by the lack of a randomly distributed exposure. And despite adjustment for the multiple potential confounders within the DCCI and APACHE II, there may still be residual confounding that contributes to the observed differences in survival. More specifically, vitamin D status may simply be a reflection of the overall health of patients, for which we may be unable to fully adjust. We were also unable to adjust for immobilization, lack of sun exposure, and vitamin D supplementation over the study period. In addition, 25OHD was only assessed within 24 h of ICU admission and may not completely represent vitamin D status at the outset of critical illness (since 25OHD levels may be influenced by inflammatory responses, fluid loading, and renal wasting of albumin as well as vitamin D binding protein) . Moreover, changes in vitamin D status over the course of the study period were not assessed. The results of our study also may not be generalizable, since patients from only one institution and from only surgical or medical ICUs were included in the analysis (i.e., excluded patients from the neurosciences, cardiac, and burn ICUs). We also had a limited sample size (n = 310), which may further impact the generalizability of our findings. Despite these limitations, the biological plausibility of vitamin D status as a determinant of survival after critical illness is undeniable. Interestingly, 25OHD levels alone had a higher AUC than the DCCI; however, it is unlikely that clinicians would be willing to make key decisions based on a single biomarker and ignore the potential impact of chronic illness on critical care outcomes. Fortunately, the addition of vitamin D status to the DCCI maintains the observed AUC of 25OHD alone and allows for a prediction model that is as accurate as APACHE II alone. As such, the results of our study warrant further investigation to confirm these findings in larger datasets and to conduct future prospective studies.
We confirm previous studies, which have suggested that total 25OHD levels upon ICU admission are inversely associated with the risk of mortality in the post-acute care setting. In addition, we present novel data that 25OHD levels on admission to the ICU may significantly improve the predictive value of the DCCI for 90-day mortality in critically ill patients. This may be particularly helpful in making clinical decisions at initiation of care in the ICU or in settings where prediction scores heavily based on physiologic assessments may not be readily available. Further prospective studies are needed to validate our findings and to determine whether optimizing vitamin D status in surgical ICU patients may confer survival benefit from critical illness.
Our Institutional Review Board proposal for this study specifically states that data will not be shared with any party outside of the Partners Healthcare Network; our database will not be available.
25OHD, 25-hydroxyvitamin D; APACHE II, Acute physiology and Chronic Health Evaluation score II; AUC, area under the curve; BMI, body mass index; DCCI, The Deyo-Charlson Comorbidity Index; EMR, Electronic medical record; ICU, intensive care unit; LOWESS, locally weighted scatterplot smoothing; MGH, Massachusetts General Hospital; OR, odds ratio; POC, point of care; ROC, receiver operating curve; SD, standard deviation; STATA, StataCorp LP; VDSP, Vitamin D Standardization Program
TMNO received support from the National Center for Advancing Translational Sciences, National Institutes of Health, Award Number TL1TR001062.
CAC received support from the National Institutes of Health grants R01 AI093723 and U01 AI087881. SAQ received support from the National Institutes of Health grants T32 GM007592, UL1 RR025758, L30 GM102903, and L30 TR001257.
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.
- Zimmerman JE, Wagner DP, Knaus WA, Williams JF, Kolakowski D, Draper EA. The use of risk predictions to identify candidates for intermediate care units. Implications for intensive care utilization and cost. Chest. 1995;108(2):490–9.PubMedView ArticleGoogle Scholar
- Kollef MH, Schuster DP. Predicting intensive care unit outcome with scoring systems. Underlying concepts and principles. Crit. Care Clin. 1994;10(1):1–18.Google Scholar
- Knaus WA, Wagner DP, Zimmerman JE, Draper EA. Variations in mortality and length of stay in intensive care units. Ann Intern Med. 1993;118(10):753–61.PubMedView ArticleGoogle Scholar
- Knaus WA, Draper EA, Wagner DP, Zimmerman JE. APACHE II: a severity of disease classification system. Crit Care Med Oct. 1985;13(10):818–29.View ArticleGoogle Scholar
- Vincent JL, Moreno R. Clinical review: scoring systems in the critically ill. Crit Care. 2010;14(2):207.PubMedPubMed CentralView ArticleGoogle Scholar
- Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis. 1987;40(5):373–83.PubMedView ArticleGoogle Scholar
- Elixhauser A, Steiner C, Harris DR, Coffey RM. Comorbidity measures for use with administrative data. Med Care. 1998;36(1):8–27.PubMedView ArticleGoogle Scholar
- Sharabiani MT, Aylin P, Bottle A. Systematic review of comorbidity indices for administrative data. Med Care. 2012;50(12):1109–18.PubMedView ArticleGoogle Scholar
- Deyo RA, Cherkin DC, Ciol MA. Adapting a clinical comorbidity index for use with ICD-9-CM administrative databases. J Clin Epidemiol. 1992;45(6):613–9.PubMedView ArticleGoogle Scholar
- Cleves MA, Sanchez N, Draheim M. Evaluation of two competing methods for calculating Charlson’s comorbidity index when analyzing short-term mortality using administrative data. J Clin Epidemiol. 1997;50(8):903–8.PubMedView ArticleGoogle Scholar
- Stukenborg GJ, Wagner DP, Connors Jr AF. Comparison of the performance of two comorbidity measures, with and without information from prior hospitalizations. Med Care. 2001;39(7):727–39.PubMedView ArticleGoogle Scholar
- Holman CD, Preen DB, Baynham NJ, Finn JC, Semmens JB. A multipurpose comorbidity scoring system performed better than the Charlson index. J Clin Epidemiol. 2005;58(10):1006–14.PubMedView ArticleGoogle Scholar
- Li P, Kim MM, Doshi JA. Comparison of the performance of the CMS Hierarchical Condition Category (CMS-HCC) risk adjuster with the Charlson and Elixhauser comorbidity measures in predicting mortality. BMC Health Serv Res. 2010;10:245.PubMedPubMed CentralView ArticleGoogle Scholar
- Chu YT, Ng YY, Wu SC. Comparison of different comorbidity measures for use with administrative data in predicting short- and long-term mortality. BMC Health Serv Res. 2010;10:140.PubMedPubMed CentralView ArticleGoogle Scholar
- Romano PS, Roos LL, Jollis JG. Adapting a clinical comorbidity index for use with ICD-9-CM administrative data: differing perspectives. J Clin Epidemiol. 1993;46(10):1075–9. discussion 1081–1090.PubMedView ArticleGoogle Scholar
- Halfon P, Eggli Y, van Melle G, Chevalier J, Wasserfallen JB, Burnand B. Measuring potentially avoidable hospital readmissions. J Clin Epidemiol. 2002;55(6):573–87.PubMedView ArticleGoogle Scholar
- Schneeweiss S, Wang PS, Avorn J, Glynn RJ. Improved comorbidity adjustment for predicting mortality in Medicare populations. Health Serv Res Aug. 2003;38(4):1103–20.View ArticleGoogle Scholar
- Quan H, Sundararajan V, Halfon P, Fong A, Burnand B, Luthi JC, Saunders LD, Beck CA, Feasby TE, Ghali WA. Coding algorithms for defining comorbidities in ICD-9-CM and ICD-10 administrative data. Med Care. 2005;43(11):1130–9.PubMedView ArticleGoogle Scholar
- Poses RM, McClish DK, Smith WR, Bekes C, Scott WE. Prediction of survival of critically ill patients by admission comorbidity. J Clin Epidemiol. 1996;49(7):743–7.PubMedView ArticleGoogle Scholar
- Fadem SZ. Charlson Comorbidity Scoring System: estimating prognosis for dialysis patients. http://touchcalc.com/calculators/cci_js. Accessed 16 Jun 2015.
- Venkatram S, Chilimuri S, Adrish M, Salako A, Patel M, Diaz-Fuentes G. Vitamin D deficiency is associated with mortality in the medical intensive care unit. Crit Care. 2011;15(6):R292.PubMedPubMed CentralView ArticleGoogle Scholar
- Braun A, Chang D, Mahadevappa K, Gibbons FK, Liu Y, Giovannucci E, Christopher KB. Association of low serum 25-hydroxyvitamin D levels and mortality in the critically ill. Crit Care Med Apr. 2011;39(4):671–7.View ArticleGoogle Scholar
- Arnson Y, Gringauz I, Itzhaky D, Amital H. Vitamin D deficiency is associated with poor outcomes and increased mortality in severely ill patients. QJM. 2012;105(7):633–9.PubMedView ArticleGoogle Scholar
- Amrein K, Quraishi SA, Litonjua AA, Gibbons FK, Pieber TR, Camargo Jr CA, Giovannucci E, Christopher KB. Evidence for a U-shaped relationship between prehospital vitamin D status and mortality: a cohort study. J Clin Endocrinol Metab. 2014;99(4):1461–9.PubMedPubMed CentralView ArticleGoogle Scholar
- Quraishi SA, Bittner EA, Blum L, McCarthy CM, Bhan I, Camargo Jr CA. Prospective study of vitamin D status at initiation of care in critically ill surgical patients and risk of 90-day mortality. Crit Care Med Jun. 2014;42(6):1365–71.View ArticleGoogle Scholar
- Joshi A, Bhadade R, Varthakavi PK, DeSouza R, Bhagwat NM, Chadha MD. Vitamin D deficiency is associated with increased mortality in critically ill patients especially in those requiring ventilatory support. Indian J Endocrinol Metab Jul. 2014;18(4):511–5.View ArticleGoogle Scholar
- Zajic P, Amrein K. Vitamin D deficiency in the ICU: a systematic review. Minerva Endocrinol. 2014;39(4):275–87.PubMedGoogle Scholar
- Padhi R, Panda B, Jagati S, Patra SC. Vitamin D status in adult critically ill patients in Eastern India: an observational retrospective study. Lung India. 2014;31(3):212–6.PubMedPubMed CentralView ArticleGoogle Scholar
- Braun AB, Gibbons FK, Litonjua AA, Giovannucci E, Christopher KB. Low serum 25-hydroxyvitamin D at critical care initiation is associated with increased mortality. Crit Care Med Jan. 2012;40(1):63–72.View ArticleGoogle Scholar
- Quraishi SA, Camargo Jr CA. Vitamin D in acute stress and critical illness. Curr Opin Clin Nutr Metab Care. 2012;15(6):625–34.PubMedPubMed CentralView ArticleGoogle Scholar
- Quraishi SA, McCarthy C, Blum L, Cobb JP, Camargo CA, Jr. Plasma 25-hydroxyvitamin D levels at initiation of care and duration of mechanical ventilation in critically ill surgical patients. JPEN J. Parenter. Enteral Nutr. 2015.Google Scholar
- Cleveland WS. Robust locally weighted regression and smoothing scatterplots. Journal of the American Statistical Association. 1979;74(368):829–36.View ArticleGoogle Scholar
- Cleveland WS, Devlin SJ. Locally weighted regression: an approach to regression analysis by local fitting. Journal of the American Statistical Association. 1988;83(403):596–610.View ArticleGoogle Scholar
- Quraishi SA, Litonjua AA, Elias KM, Gibbons FK, Giovannucci E, Camargo Jr CA, Christopher KB. Association between pre-hospital vitamin D status and hospital-acquired new-onset delirium. Br J Nutr. 2015;113(11):1753–60.PubMedPubMed CentralView ArticleGoogle Scholar
- Quraishi SA, Litonjua AA, Moromizato T, Gibbons FK, Camargo Jr CA, Giovannucci E, Christopher KB. Association between prehospital vitamin D status and hospital-acquired Clostridium difficile infections. JPEN J Parenter Enteral Nutr Jan. 2015;39(1):47–55.View ArticleGoogle Scholar
- Quraishi SA, Litonjua AA, Moromizato T, Gibbons FK, Camargo Jr CA, Giovannucci E, Christopher KB. Association between prehospital vitamin D status and hospital-acquired bloodstream infections. Am J Clin Nutr. 2013;98(4):952–9.PubMedPubMed CentralView ArticleGoogle Scholar
- Cleves MA. From the help desk: comparing areas under receiver operating characteristic curves from two or more probit or logit models. The Stata Journal. 2002;2(3):301–13.Google Scholar
- Ho KM, Dobb GJ, Knuiman M, Finn J, Lee KY, Webb SA. A comparison of admission and worst 24-hour Acute Physiology and Chronic Health Evaluation II scores in predicting hospital mortality: a retrospective cohort study. Crit Care. 2006;10(1):R4.PubMedView ArticleGoogle Scholar
- Capuzzo M, Valpondi V, Sgarbi A, Bortolazzi S, Pavoni V, Gilli G, Candini G, Gritti G, Alvisi R. Validation of severity scoring systems SAPS II and APACHE II in a single-center population. Intensive Care Med. 2000;26(12):1779–85.PubMedView ArticleGoogle Scholar
- Shen JJ, Ginn GO. Financial position and adoption of electronic health records: a retrospective longitudinal study. J Health Care Finance Spring. 2012;38(3):61–77.Google Scholar
- Quach S, Hennessy DA, Faris P, Fong A, Quan H, Doig C. A comparison between the APACHE II and Charlson Index Score for predicting hospital mortality in critically ill patients. BMC Health Serv Res. 2009;9:129.PubMedPubMed CentralView ArticleGoogle Scholar
- Govani SM, Waljee AK, Stidham RW, Higgins PD. Increasing ultraviolet light exposure is associated with reduced mortality from Clostridium difficile infection. United European gastroenterology journal. 2015;3(2):208–14.PubMedPubMed CentralView ArticleGoogle Scholar
- Quan H, Li B, Couris CM, Fushimi K, Graham P, Hider P, Januel JM, Sundararajan V. Updating and validating the Charlson comorbidity index and score for risk adjustment in hospital discharge abstracts using data from 6 countries. Am J Epidemiol. 2011;173(6):676–82.PubMedView ArticleGoogle Scholar
- Christensen S, Johansen MB, Christiansen CF, Jensen R, Lemeshow S. Comparison of Charlson comorbidity index with SAPS and APACHE scores for prediction of mortality following intensive care. Clin Epidemiol. 2011;3:203–11.PubMedPubMed CentralView ArticleGoogle Scholar
- Moromizato T, Litonjua AA, Braun AB, Gibbons FK, Giovannucci E, Christopher KB. Association of low serum 25-hydroxyvitamin D levels and sepsis in the critically ill. Crit Care Med Jan. 2014;42(1):97–107.View ArticleGoogle Scholar
- Amrein K, Venkatesh B. Vitamin D and the critically ill patient. Curr Opin Clin Nutr Metab Care. 2012;15(2):188–93.PubMedView ArticleGoogle Scholar
- Maiya S, Sullivan I, Allgrove J, Yates R, Malone M, Brain C, Archer N, Mok Q, Daubeney P, Tulloh R, Burch M. Hypocalcaemia and vitamin D deficiency: an important, but preventable, cause of life-threatening infant heart failure. Heart. 2008;94(5):581–4.PubMedView ArticleGoogle Scholar
- Bouillon R, Carmeliet G, Verlinden L, van Etten E, Verstuyf A, Luderer HF, Lieben L, Mathieu C, Demay M. Vitamin D and human health: lessons from vitamin D receptor null mice. Endocr Rev. 2008;29(6):726–76.PubMedPubMed CentralView ArticleGoogle Scholar
- Bischoff-Ferrari HA, Conzelmann M, Dick W, Theiler R, Stahelin HB. [Effect of vitamin D on muscle strength and relevance in regard to osteoporosis prevention]. Z. Rheumatol. 2003;62(6):518–21.Google Scholar
- Kankova M, Luini W, Pedrazzoni M, Riganti F, Sironi M, Bottazzi B, Mantovani A, Vecchi A. Impairment of cytokine production in mice fed a vitamin D3-deficient diet. Immunology. 1991;73(4):466–71.PubMedPubMed CentralGoogle Scholar
- Pinheiro Da Silva F, Machado MC. Antimicrobial peptides: clinical relevance and therapeutic implications. Peptides. 2012;36(2):308–14.PubMedView ArticleGoogle Scholar
- Durr UH, Sudheendra US, Ramamoorthy A. LL-37, the only human member of the cathelicidin family of antimicrobial peptides. Biochim Biophys Acta. 2006;1758(9):1408–25.PubMedView ArticleGoogle Scholar
- Tollin M, Bergman P, Svenberg T, Jornvall H, Gudmundsson GH, Agerberth B. Antimicrobial peptides in the first line defence of human colon mucosa. Peptides. 2003;24(4):523–30.PubMedView ArticleGoogle Scholar
- Fabri M, Stenger S, Shin DM, Yuk JM, Liu PT, Realegeno S, Lee HM, Krutzik SR, Schenk M, Sieling PA, Teles R, Montoya D, Iyer SS, Bruns H, Lewinsohn DM, Hollis BW, Hewison M, Adams JS, Steinmeyer A, Zugel U, Cheng G, Jo EK, Bloom BR, Modlin RL. Vitamin D is required for IFN-gamma-mediated antimicrobial activity of human macrophages. Sci Transl Med. 2011;3(104):104ra102.PubMedPubMed CentralView ArticleGoogle Scholar
- Quraishi SA, De Pascale G, Needleman JS, Nakazawa H, Kaneki M, Bajwa EK, Camargo Jr CA, Bhan I. Effect of cholecalciferol supplementation on vitamin D status and cathelicidin levels in sepsis: a randomized placebo-controlled trial. Crit Care Med Sep. 2015;43(9):1928–37.View ArticleGoogle Scholar
- Asakura H, Aoshima K, Suga Y, Yamazaki M, Morishita E, Saito M, Miyamoto K, Nakao S. Beneficial effect of the active form of vitamin D3 against LPS-induced DIC but not against tissue-factor-induced DIC in rat models. Thromb Haemost. 2001;85(2):287–90.PubMedGoogle Scholar
- Moller S, Laigaard F, Olgaard K, Hemmingsen C. Effect of 1,25-dihydroxy-vitamin D3 in experimental sepsis. Int J Med Sci. 2007;4(4):190–5.PubMedPubMed CentralView ArticleGoogle Scholar
- Horiuchi H, Nagata I, Komoriya K. Protective effect of vitamin D3 analogues on endotoxin shock in mice. Agents Actions. 1991;33(3–4):343–8.PubMedView ArticleGoogle Scholar
- Vitamin D. Metabolites quality assurance program. 2010. http://www.nist.gov/mml/csd/vitdqap.cfm. Accessed 30 May 2016.Google Scholar
- Tai SS, Bedner M, Phinney KW. Development of a candidate reference measurement procedure for the determination of 25-hydroxyvitamin D3 and 25-hydroxyvitamin D2 in human serum using isotope-dilution liquid chromatography-tandem mass spectrometry. Anal Chem. 2010;82(5):1942–8.PubMedPubMed CentralView ArticleGoogle Scholar
- Stepman HC, Vanderroost A, Van Uytfanghe K, Thienpont LM. Candidate reference measurement procedures for serum 25-hydroxyvitamin D3 and 25-hydroxyvitamin D2 by using isotope-dilution liquid chromatography-tandem mass spectrometry. Clin Chem. 2011;57(3):441–8.PubMedView ArticleGoogle Scholar
- Stockl D, Sluss PM, Thienpont LM. Specifications for trueness and precision of a reference measurement system for serum/plasma 25-hydroxyvitamin D analysis. Clin Chim Acta. 2009;408(1–2):8–13.PubMedView ArticleGoogle Scholar
- Lee S, Oncescu V, Mancuso M, Mehta S, Erickson D. A smartphone platform for the quantification of vitamin D levels. Lab on a chip. 2014;14(8):1437–42.PubMedView ArticleGoogle Scholar
- The Fast Pack System 2013. www.qualigeninc.com. Accessed 30 May 2016.
- Point of Care Diagnostics 2015. http://www.nanospeed.ca/. Accessed 30 May 2016.