Collaborative and privacy-enhancing workflows on a clinical data warehouse: an example developing natural language processing pipelines to detect medical conditions.

OBJECTIVE: To develop and validate a natural language processing (NLP) pipeline that detects 18 conditions in French clinical notes, including 16 comorbidities of the Charlson index, while exploring a collaborative and privacy-enhancing workflow. MATERIALS AND METHODS: The detection pipeline relied both on rule-based and machine learning algorithms, respectively, for named entity recognition and entity qualification, respectively. We used a large language model pre-trained on millions of clinical notes along with annotated clinical notes in the context of 3 cohort studies related to oncology, cardiology, and rheumatology. The overall workflow was conceived to foster collaboration between studies while respecting the privacy constraints of the data warehouse. We estimated the added values of the advanced technologies and of the collaborative setting. RESULTS: The pipeline reached macro-averaged F1-score positive predictive value, sensitivity, and specificity of 95.7 (95%CI 94.5-96.3), 95.4 (95%CI 94.0-96.3), 96.0 (95%CI 94.0-96.7), and 99.2 (95%CI 99.0-99.4), respectively. F1-scores were superior to those observed using alternative technologies or non-collaborative settings. The models were shared through a secured registry. CONCLUSIONS: We demonstrated that a community of investigators working on a common clinical data warehouse could efficiently and securely collaborate to develop, validate and use sensitive artificial intelligence models. In particular, we provided an efficient and robust NLP pipeline that detects conditions mentioned in clinical notes.

Risk factors associated with bacteremia in COVID-19 patients admitted to intensive care unit: a retrospective multicenter cohort study.

PURPOSE: This multicenter observational study was done to evaluate risk factors related to the development of BSI in patients admitted to ICU for COVID-19. METHODS: All patients with COVID-19 admitted in two COVID-19 dedicated ICUs in two different hospital between 02-2020 and 02-2021 were recruited. RESULT: 537 patients were included of whom 265 (49.3%) experienced at least one BSI. Patients who developed bacteremia had a higher SOFA score [10 (8-12) vs 9 (7-10), p < 0.001], had been intubated more frequently [95.8% vs 75%, p < 0.001] and for a median longer time [16 days (9-25) vs 8 days (5-14), p < 0.001]. Patients with BSI had a median longer ICU stay [18 days (12-31.5) vs 9 days (5-15), p < 0.001] and higher mortality [54% vs 42.3%, p < 0.001] than those who did not develop it. Development of BSI resulted in a higher SOFA score [aHR 1.08 (95% CI 1.03-1.12)] and a higher Charlson score [csAHR 1.15 (95% CI 1.05-1.25)]. CONCLUSION: A high SOFA score and a high Charlson score resulted associated with BSI's development. Conversely, immunosuppressive therapy like steroids and tocilizumab, has no role in increasing the risk of bacteremia.

Elixhauser outperformed Charlson comorbidity index in prognostic value after ACS: insights from a national registry.

OBJECTIVE: To compare the performance of risk adjustment models using the Elixhauser and Charlson comorbidity scores in predicting in-hospital outcomes of ACS patients from a nationwide administrative database. STUDY DESIGN AND SETTING: All hospitalizations for ACS in the United States between 2004 and 2014 (n = 7,201,900) were retrospectively analyzed. We used ECS and CCI score based on ICD-9 codes to define comorbidity variables. Logistic regression models were fitted to three in-hospital outcomes, including mortality, Major Acute Cardiovascular & Cerebrovascular Events (MACCE) and bleeding. The prognostic values of ECS and CCI after adjusting for known confounders, were compared using the C-statistic, Akaike information criterion (AIC), and Bayesian information criterion (BIC). RESULTS: The statistical performance of models predicting all in-hospital outcomes demonstrated that the ECS had superior prognostic value compared to the CCI, with higher C-statistics and lower AIC and BIC values associated with the former. CONCLUSION: This is the first study that compared the prognostic value of the ECS and CCI scores in predicting multiple ACS outcomes, based on their scoring systems. Better discrimination and goodness of fit was achieved with the Elixhauser method across all in-hospital outcomes studied.

Antibiotics prescription and guidelines adherence in elderly: impact of the comorbidities.

BACKGROUND: Although the interest of antibiotics is well known, antibiotics prescription is associated with side effect, especially in patients with multiples comorbidities. One way to reduce the incidence of side effects is to respect antibiotics prescriptions guidelines. Our objective was to investigated the factors associated with guidelines adherence in elderly patients with multiples comorbidities. METHODS: From October 2015 to December 2016, antibiotics prescription and guidelines adherence were analyzed in two post-acute care and rehabilitation services of a 2600-bed, university-affiliated center. RESULTS: One hundred and twenty-eight patients were included, fifty-nine (46%) patients had antibiotics prescription according to guidelines. In Multivariable logistic regression analysis, prescription of 2 antibiotics or more (OR = 0.168, 95% IC = 0.037-0.758, p < 0.05), 85 years of age and more (OR = 0.375, 95% IC = 0.151-0.931, p < 0.05) and the Charlson comorbidity index score (OR = 0.750, 95% IC = 0.572-0.984, p < 0.05) were negatively associated with antibiotics prescriptions according to guidelines. CONCLUSIONS: High comorbidity in the elderly was negatively associated with the guidelines adherence of antibiotiсs prescriptions. These criteria should be considered to optimize antibiotics prescriptions in elderly patients.

The burden of frailty in older people visiting GPs in Veneto and Sicily, Italy.

Context: In Italy, little is known about the territorial distribution of the frailty status. Aims: To compare frailty- and multimorbidity-prevalence in the elderly population of two Italian regions. Methods: This study examined randomized samples of elderly (both community dwelling and institutionalized) assisted by general practitioners. Frailty was evaluated through the CSHA-Scale, multimorbidity through the Charlson-Score. The relation between frailty and multimorbidity was studied through a logistic model. Both crude and standardized prevalences were calculated. Results: One hundred and sixteen physicians assisted 176,503 patients highly representative of Italian people. In a randomized sample of 4,531 older people, the sex-age-standardized prevalence of Frailty (standard population: Italy) was 25.74% (24.63-26.85%). Age-standardized prevalence for males was 20.08% (18.46-21.71%) and 30.00% (28.54-31.57%) for females. Using the sex-age-standardization pooled sample, the prevalence of frailty was significantly higher in Sicily than Veneto (28.74% [27.03-30.46%] vs 22.30% [20.94-23.67%]. This study did not find differences in the prevalence of multimorbidity: Veneto 20.76% (19.21-22.31%); Sicily 22.05% (20.33-23.77%). Both "to be female" and "to live in Sicily" were shown to be predictors of frailty OR for being female = 1.64 (1.42-1.88); OR for living in Sicily = 1.27 (1.11-1.46). Multimorbidity was an independent frailty-predictor only for those aged < 85: OR of Charlson Index ≥ 4 for ages < 85 = 3.44 (2.88-4.11), OR for ages ≥ 85 = 1.44 (0.97-2.12). Limitations: (1) This study considered patients assisted by doctors, not a random sample of the general population. (2) The cross-sectional nature of the study limits the interpretation of the relationships between frailty and multi-morbidity. (3) Few covariates were available for our multivariate models. Conclusions: More than 1/4 of elderly persons are shown to be frail (1/5 of males and 1/3 of females). Frailty is more frequent in Sicily, while multimorbidity does not differ between the two regions. This could be due to regional differences in the organization of care networks dedicated to elderly patients.

Risk factors and outcomes of Clostridium difficile infection in hospitalized patients.

BACKGROUND: The aim of this study was to identify risk factors for Clostridium difficile infection (CDI) and its attributable mortality and to propose methods to prevent CDI and improve patients' outcomes. METHODS: CDI was defined as diarrheal patients with stool samples that were positive for C. difficile toxin. Clinical presentations of all patients with CDI and two times as many age- and sex-matched culture-negative controls at the Chang Gung Memorial Hospital in 2014 were identified and compared by multivariate, nonparametric, and Kaplan-Meier survival analysis. RESULTS: There were no significant differences in ages, sex, or Charlson comorbidity indexes between the CDI group (n = 42) and the control group (n = 86). The multivariate analysis indicated that underlying peptic ulcer disease and previous use of gastric acid inhibitors or third-generation cephalosporins for at least 3 days were significantly more common in patients with CDI than in the controls. Charlson scores were associated with mortality due to CDI. Recommended treatment using oral vancomycin to treat patients with Charlson score ≥ 5 and oral metronidazole or vancomycin to treat those with moderate underlying disease (Charlson score ≥ 2 and ≤ 5) significantly increased survival in these patients (p = 0.001). CONCLUSIONS: Oral vancomycin given to patients with high Charlson scores and oral metronidazole or vancomycin to patients with moderate Charlson scores decreased mortality due to CDI. Restricting the use of third-generation cephalosporins and gastric acid inhibitors is recommended to prevent CDI in hospitalized patients.

Acute pancreatitis in elderly patients: a single-center retrospective evaluation of clinical outcomes.

Introduction: Acute pancreatitis (AP) incidence in the elderly population has increased in the last years. However, the role of age as influencing factor on the AP clinical course is still debated. Methods: We reviewed clinical records of consecutive patients admitted with diagnosis of AP. Patients were divided in elderly (≥65 years) and non-elderly (<65 years). Primary endpoint was comparison of overall mortality. Secondary endpoint included ICU admission, in-hospital length of stay (LOS) and surgical procedures. Results: We enrolled 352 elderly and 532 non-elderly patients. A higher mortality rate (7.4% vs 1.9%; p < .001), ICU admission rate (18.9% vs 6.3%; p < .001) and prolonged length of hospital stay (9 (6-14) vs 7 (5-11.7) days; p = .01) were registered in the ≥65 years group. Multivariate analysis identified age (OR: 3.5; 95% CI:1.645-7.555; p = .001), a higher Ranson score at admission (OR: 5.52; 95% CI:1.11-27.41; p<.001) and necrotic pancreatitis (OR: 8.6; 95% CI:2.46-30.27; p = .001) as independent predictors of mortality. Conversely age and necrotic pancreatitis were independent risk factors for higher LOS and ICU admission. Conclusions: Patients with AP and age ≥65 years have a higher mortality, ICU admission and prolonged LOS. Early recognition and prompt treatment are key elements to improve outcomes in this population.

Prevalence of frailty in older people in Veneto (Italy).

Context: Both frailty and multimorbidity are strong predictors of clinical endpoints for older people. In Italy, the interventions targeting chronicity are mainly based on the treatment of diseases: sufficient epidemiological literature is available about these strategies. Less is known about the territorial distribution of the frailty status. Aims: To estimate the prevalence of frailty in older people (65+) and to evaluate the relationship between frailty and multimorbidity. Methods and material: A group of general practitioners working in Veneto (Italy) was enrolled on a voluntary basis. Older individuals were both community dwelling and institutionalized patients, that is, the older people normally followed by Italian general practitioners. A centrally randomized sample was extracted from the pool of physician-assisted elderly. Each doctor evaluated the frailty status through the CSHA Clinical Frailty Scale and the multimorbidity status through the Charlson score (Frailty = CSHA Clinical Frailty Scale's score >4; serious multimorbidity = Charlson score ≥4). Prevalence and its confidence interval (CI) 95% were evaluated through the Agresti's method for proportions. The relation between frailty and multimorbidity was studied through a logistic regression model adjusted for age and sex. Results: Fifty-three physicians were enrolled, whose population of elderly individuals (N = 82919) was highly representative of the population of Veneto. The prevalence of frailty in the randomized sample of 2407 older people was 23.18% (CI 95%: 21.53%-24.91%). Sex was shown to be a strong predictor of frailty (female status OR = 1.58 p < .0001) and multimorbidity was shown to be an independent predictor only for individuals <85 years of age. Conclusions: In Veneto, more than 20% of elderly people are frail. Physicians should pay close attention to frailty and multimorbidity because both are important prognostic factors toward clinical endpoints relevant to territorial care. The CSHA Clinical Frailty Scale (easy and quick) should become part of their professional routine.

The relationship of Charlson comorbidity index with stent restenosis and extent of coronary artery disease.

OBJECTIVES: The objective of this study is to investigate the effect of comorbid conditions [Charlson comorbidity index (CCI)] on stent restenosis who underwent coronary angioplasty earlier. METHODS: Patients were divided into two groups; patients with critical restenosis [recurrent diameter stenosis >50% at the stent segment or its edges (5-mm segments adjacent to the stent) (Group 1; n = 53, mean age: 63.8 ± 9.9 years)] and patients with no critical restenosis [<50% obstruction (Group 2; n = 94, mean age: 62.1 ± 9.1 years)]. The CCI and modified CCI were used for the presence of comorbid conditions. The Gensini scoring system was used to assess the extent of coronary artery disease (CAD). RESULTS: Group 1 had a significantly greater CCI and modified CCI score compared to Group 2 (7.1 ± 3.7 vs. 5.6 ± 1.6, p = 0.006; 6.9 ± 3.6 vs. 4.5 ± 1.5, p = 0.008, respectively). There was a weak correlation, albeit significant, between the modified CCI score and restenosis percentage (r = 0.29, p < 0.001; r = 0.25, p = 0.003, respectively). CONCLUSIONS: In conclusion, the CCI score is greater among patients with stent restenosis than those without. CCI score is higher among patients with a more diffuse CAD than with a milder disease extent.

Influence of age and comorbidity on prognosis and application of adjuvant chemotherapy in elderly Japanese patients with colorectal cancer: A retrospective multicentre study.

BACKGROUND: Adjuvant therapy for colorectal cancer (CRC) in patients aged ≥75 years is supported by inadequate evidence, although such patients are increasing in number worldwide. PATIENTS AND METHODS: We assessed the influence of age and comorbidities on the prognosis of CRC in elderly patients using pooled data by the Japanese Study Group for Postoperative Follow-up of Colorectal Cancer. In total, 4598 patients (3304 with colon cancer and 1294 with rectal cancer) who underwent curative surgery from 2004 to 2006 were analysed with respect to age, Charlson comorbidity score (CS), tumour marker positivity, adjuvant therapy and prognosis. RESULTS: The number of patients aged <64, 65-74 and >75 years was 2007 (44%), 1614 (35%) and 977 (21%), respectively. Tumour location, tumour marker positivity, clinical stage, performance of adjuvant therapy, CS and overall survival (OS) were significantly different among these age groups (P < 0.0001). Among patients aged ≥75 years with stage III CRC, 35% with colon cancer and 21% with rectal cancer received adjuvant therapy; these proportions were much lower than those in younger patients. Application of adjuvant therapy was dependent on the CS in patients aged ≤74 years, but not in older patients. Sex, the carcinoembryonic antigen concentration and adjuvant therapy were significantly associated with OS in elderly patients with stage III CRC. CONCLUSION: Age and comorbidities worsened the OS of patients with CRC who underwent curative surgery. However, patients aged ≥75 years were undertreated regardless of their CS despite the possibility of OS improvement by adjuvant therapy.

Locally advanced breast cancer in elderly patients.

BACKGROUND: Although locally advanced breast cancer (LABC) is more common in the elderly population, there is little data on the clinical characteristics and survival of these patients. The aim of the present study was to compare different factors affecting survival in elderly patients with LABC. METHODS: Retrospective analysis was carried out on a cohort of 80 patients aged 70 to 96 years, diagnosed with LABC defined as T3 N1, T4 N0, any N2 or N3, and M0. The prognostic impact of selected clinical parameters including age, comorbidities, tumour grade, HER2 status, tumour stage, local therapies, and systemic treatments was studied. RESULTS: The median age of the patients was 79 years. The majority (n=53; 66%) had at least one significant comorbidity according to the Charlson score evaluation. The median overall survival was 50.6 months. As expected, hormonal therapy was the dominant mode of systemic treatment, but 24% also received at least one line of chemotherapy. Local therapies including surgery and/or radiotherapy were applied in 58% of patients. CONCLUSIONS: The diagnosis of LABC in the elderly is associated with poor prognosis. Age and serious comorbidities were negative prognostic factors.

Current impact of age and comorbidity assessment on prostate cancer treatment choice and over/undertreatment risk.

PURPOSE: We evaluated the influence of age and comorbidity (Charlson score assessment) on localized prostate cancer therapeutic management and the risk of prostate cancer over- and under-treatment. METHODS: Among the 2571 prostate cancer cases diagnosed in 2011, a subset of 633 patients was randomly selected from the prospectively accrued cohort of the Regional Cancer Registry, among the 17 participating institutions. Treatment distributions were examined for patients at each individual prostate cancer risk, age and comorbidity level and analyzed by multivariate logistic regression analysis. RESULTS: Treatments with curative intent were observed less often when age increased (p < 0.001). We found no impact of the Charlson score on the selection of a curative treatment [HR 0.89, 95 % CI (0.70-1.15)]. A 20 % overtreatment rate was reported in low-risk prostate cancer patients. For younger patients (65-75 years) with high comorbidity score, a 14 % overtreatment rate was observed. Conversely, a 16 % undertreatment rate was reported in older patients >75 years without any significant comorbidity. CONCLUSION: A better consideration of comorbidities could significantly reduce overtreatment in patients <75 year and promote curative treatment in aggressive prostate cancer for older patients without any significant comorbidity.

Outcome of rectal cancer after radiotherapy with a long or short waiting period before surgery, a descriptive clinical study.

BACKGROUND: Radiotherapy and surgery have shown to improve local control and survival in rectal cancer. There are two applied schedules; radiotherapy with a long or short waiting period before surgery. The effect on survival and recurrence of both schedules was studied. METHODS: All consecutive patients with rectal cancer in the period 2002-2008 were included. Data were gathered on survival, tumour stage, co-morbidity score, and cause of death. The patients were divided in three groups: group 1 patients undergoing surgery without neo-adjuvant radiotherapy; group 2 patients undergoing radiotherapy followed by immediate surgery; and group 3 patients treated with (chemo) radiotherapy followed by a longer waiting period. RESULTS: A total of 113 patients with rectal cancer underwent surgery. Twenty two patients in group 1, 71 patients in group 2, and 20 in group 3. There was no difference in gender, time to recurrence, co-morbidity score, or causes of death. Fifty percent of patients died due to non-cancer related causes. Mean age in patients of group 3 was significantly lower than in groups 1 and 2 (P=0.02). There was a trend towards a lower tumour stage in the patients of group 3. Overall five year survival was 32% in group 1, 48% in group 2, and 35% in group 3. CONCLUSIONS: Neo-adjuvant radiotherapy seems to be of benefit in daily practice in patients with rectal cancer. A longer waiting period results in down-staging. Clinicians have to be aware that many patients will die due to other causes than those related to the rectal cancer.

Impact of age and comorbidity on survival in colorectal cancer.

BACKGROUND: Patients with colorectal cancer are often excluded from clinical trials based on age or a poor performance score. However, 70% of colorectal cancer is diagnosed in patients over 65. Evaluation on the influence of age and comorbidity on survival and cause of death in a non-selected population. METHODS: Included were 621 consecutive patients with colorectal cancer. An extensive chart review was performed for 392 patients with colon cancer and 143 patients with rectal cancer. Analyses were performed separately for both groups. RESULTS: Median survival of colon cancer patients was 5.13 years, 131 patients (34.3%) died from tumour progression. Age and comorbidity were significant predictors for overall survival (P<0.001). Age was also a significant predictor of cause of death (P=0.001). In rectal cancer patients median survival was 4.67 years, 51 (35.7%) of patients died from tumour progression. Neither age nor comorbidity was significant predictors of survival. Age was a significant predictor of cause of death (P<0.001). CONCLUSIONS: In colon cancer patient age and comorbidity predict survival. This represents possible bias or a reduced survival benefit of treatment, and is an indication that colon cancer is not the prognosis defining illness in the majority of patients. In rectal cancer patients neither age or comorbidity significantly impacted survival.

Thirty days post-operative mortality after surgery for colorectal cancer: a descriptive study.

BACKGROUND: The goal of surgery for colorectal cancer is cure. Unfortunately post-operative mortality occurs. This study aims to identify co-morbidity and causes of mortality in the post-operative period in relation to direct technical complications of surgery. METHODS: All consecutive patients who underwent surgery for colorectal cancer were included. Co-morbidity was determined via the Charlson co-morbidity score. The post-operative course was studied and cause of death within 30 days was determined. Patients were divided in two groups: group 1 died within 30 days after surgery and group 2 survived for longer than 30 days. RESULTS: Twenty three out of 333 patients (6.9%) with colon cancer and 6 out of 112 (5.3%) with rectal cancer died in the post-operative period. Patients in group 1 were significantly older than patients in group 2 (P<0.001). Patients in group 1 with colon cancer also significantly had more often a higher stage of cancer (P=0.03). The Charlson co-morbidity score for patients with colon cancer in group 1 was mean 5.17 (SD 1.57, range, 1-8), and for rectal cancer mean 4.83 (SD 2.32, range, 2-7). There was no difference in Charlson co-morbidity score when patients from groups 1 and 2 were compared. In group 1, 13 (44%) died as a direct consequence of technical surgical complications. Sixteen patients died due to complications because of pre-existing co-morbidity. CONCLUSIONS: Post-operative mortality very often is the direct result of pre-existing co-morbidity and not always the direct result of the surgical procedure.

An evaluation of the Charlson co-morbidity score for predicting sepsis after elective major surgery.

BACKGROUND AND AIMS: Severe sepsis is a significant cause of morbidity and mortality following major surgery. The Charlson co-morbidity score (CCS) has been shown to be associated with severe sepsis following major surgery for cancer. This prospective observational study investigated the effect of patient factors (CCS, gender, age and malignancy) and intraoperative factors (duration of surgery and allogeneic blood transfusion) on the incidence of sepsis after elective major surgery, and the impact of patient co-morbidities on length of stay in critical care. MATERIALS AND METHODS: We prospectively identified a cohort of 101 patients undergoing elective major surgery in a university teaching hospital. The CCS was calculated before surgery, and the incidence of sepsis was documented following surgery. We investigated whether age, malignancy, intraoperative allogeneic blood transfusion, length of surgery or gender were associated with sepsis following surgery. RESULTS: Twenty-seven (27%) patients developed sepsis. Using multivariate logistic regression, the duration of surgery was associated with the development of sepsis after surgery (P = 0.054, odds ratio 1.2). The CCS was not associated with sepsis in this population of cancer and non-cancer patients undergoing elective major surgery, but was associated with longer length of stay in the intensive care unit (P = 0.016). CONCLUSIONS: Duration of surgery, but not patient co-morbidity as assessed by the CCS, may predict the postoperative incidence of sepsis. CCS could be used as a guide to predict consumption of critical care resources by elective surgical patients. A higher CCS was associated with a longer ICU stay. Resources, such as postoperative goal directed therapy, may be useful in reducing length of stay, hospital costs and risks of infective complications in this subgroup of patients with higher CCS.