Evaluating statistical and machine learning methods to predict risk of in-hospital child mortality in Uganda

Thesis (Master's)--University of Washington, 2016-08

Excessive blood pressure increase with exercise and risk of all-cause mortality and cardiac events

The association of an excessive blood pressure increase with exercise (EBPIE) on cardiovascular outcomes remains controversial. We sought to assess its impact on the risk of all-cause mortality and major cardiac events in patients with known or suspected coronary artery disease (CAD) referred for stress testing. Exercise echocardiography was performed in 10,047 patients with known or suspected CAD. An EBPIE was defined as an increase in systolic blood pressure with exercise ≥80 mmHg. The endpoints were all-cause mortality and major cardiac events (MACE), including cardiac death or nonfatal myocardial infarction (MI). Overall, 573 patients exhibited an EBPIE during the tests. Over a mean follow-up of 4.8 years, there were 1,950 deaths (including 725 cardiac deaths), 1,477 MI, and 1,900 MACE. The cumulative 10-year rates of all-cause mortality, cardiac death, nonfatal MI and MACE were 32.9%, 13.1%, 26,9% and 33% in patients who did not develop an EBPIE vs. 18.9%, 4.7%, 17.5% and 20.7% in those experiencing an EBPIE, respectively (p <0.001 for all comparisons). In Cox regression analyses, an EBPIE remained predictive of all-cause mortality (hazard ratio [HR] 0.73, 95% confidence interval [CI] 0.59-0.91, p = 0.004), cardiac death (HR 0.67, 95% CI 0.46-0.98, p = 0.04), MI (HR 0.67, 95% CI 0.52-0.86, p = 0.002), and MACE (HR 0.69, 95% CI 0.56-0.86, p = 0.001). An EBPIE was associated with a significantly lower risk of mortality and MACE in patients with known or suspected CAD referred for stress testing.

Prevalence and risk factors for aminoglycoside nephrotoxicity in intensive care units

In order to assess the prevalence of and risk factors for aminoglycoside-associated nephrotoxicity in intensive care units (ICUs), we evaluated 360 consecutive patients starting aminoglycoside therapy in an ICU. The patients had a baseline calculated glomerular filtration rate (cGFR) of ?30 ml/min/1.73 m2. Among these patients, 209 (58 per cent) developed aminoglycoside-associated nephrotoxicity (the acute kidney injury [AKI] group, which consisted of individuals with a decrease in cGFR of >20 per cent from the baseline cGFR), while 151 did not (non-AKI group). Both groups had similar baseline cGFRs. The AKI group developed a lower cGFR nadir (45 ± 27 versus 79 ± 39 ml/min/1.73 m2 for the non-AKI group; P < 0.001); was older (56 ± 18 years versus 52 ± 19 years for the non-AKI group; P = 0.033); had a higher prevalence of diabetes (19.6 per cent versus 9.3 per cent for the non-AKI group; P = 0.007); was more frequently treated with other nephrotoxic drugs (51 per cent versus 38 per cent for the non-AKI group; P = 0.024); used iodinated contrast more frequently (18 per cent versus 8 per cent for the non-AKI group; P = 0.0054); and showed a higher prevalence of hypotension (63 per cent versus 44 per cent for the non-AKI group; P = 0.0003), shock (56 per cent versus 31 per cent for the non-AKI group; P < 0.0001), and jaundice (19 per cent versus 8 per cent for the non-AKI group; P = 0.0036). The mortality rate was 44.5 per cent for the AKI group and 29.1 per cent for the non-AKI group (P = 0.0031). A logistic regression model identified as significant (P < 0.05) the following independent factors that affected aminoglycoside-associated nephrotoxicity: a baseline cGFR of <60 ml/min/1.73 m2 (odds ratio [OR], 0.42), diabetes (OR, 2.13), treatment with other nephrotoxins (OR, 1.61) or iodinated contrast (OR, 2.13), and hypotension (OR, 1.83). (To continue) In conclusion, AKI was frequent among ICU patients receiving an aminoglycoside, and it was associated with a high rate of mortality. The presence of diabetes or hypotension and the use of other nephrotoxic drugs and iodinated contrast were independent risk factors for the development of aminoglycoside-associated nephrotoxicity

Estimating cardiovascular risk in patients with type 2 diabetes: a national multicenter study in Brazil

According to Brazilian National Data Survey diabetes is the fifth cause for hospitalization and is one of the ten major causes of mortality in this country. Aims to stratify the estimated cardiovascular risk (eCVR) in a population of type 2 diabetics (T2DM) according to the Framingham prediction equations as well as to determine the association between eCVR with metabolic and clinical control of the disease. Methods From 2000 to 2001 a cross-sectional multicenter study was conducted in 13 public out-patients diabetes/endocrinology clinics from 8 Brazilian cities. The 10-year risk of developing coronary heart disease (CHD) was estimated by the prediction equations described by Wilson et al (Circulation 1998). LDL equations were preferably used; when patients missed LDL data we used total cholesterol equations instead. Results Data from 1382 patients (59.0% female) were analyzed. Median and inter-quartile range (IQ) of age and duration of diabetes were 57.4 (51-65) and 8.8 (3-13) years, respectively without differences according to the gender. Forty-two percent of these patients were overweight and 35.4% were obese (the prevalence of higher BMI and obesity in this T2DM group was significantly higher in women than in men; p < 0.001). The overall estimated eCVR in T2DM patients was 21.4 (13.5-31.3). The eCVR was high (> 20%) in 738 (53.4%), intermediate in 202 (14.6%) and low in 442 (32%) patients. Men [25.1(15.4-37.3)] showed a higher eCVR than women [18.8 (12.4-27.9) p < 0.001]. The most common risk factor was high LDL-cholesterol (80.8%), most frequently found in women than in men (p = 0.01). The median of risk factors present was three (2-4) without gender differences. Overall we observed that 60 (4.3%) of our patients had none, 154(11.1%) one, 310 (22.4%) two, 385 (27.9%) three, 300 (21.7%) four, 149 (10.5%) five and six, (2%) six risk factors. A higher eCVR was noted in overweight or obese patients (p = 0.01 for both groups). No association was found between eCVR with age or a specific type of diabetes treatment. A correlation was found between eCVR and duration of diabetes (p < 0.001), BMI (p < 0.001), creatinine (p < 0.001) and triglycerides levels (p < 0.001) but it was not found with HbA1c, fasting blood glucose and postprandial glucose. A higher eCVR was observed in patients with retinopathy (p < 0.001) and a tendency in patients with microalbuminuria (p = 0.06). Conclusion: our study showed that in this group of Brazilian T2DM the eCVR was correlated with the lipid profile and it was higher in patients with microvascular chronic complications. No correlation was found with glycemic control parameters. These data could explain the failure of intensive glycemic control programs aiming to reduce cardiovascular events observed in some studies.

TCF7L2 Polymorphism rs7903146 Is Associated with Coronary Artery Disease Severity and Mortality

Background: TCF7L2 polymorphisms have been consistently associated with type 2 diabetes mellitus in different populations and type 2 diabetes mellitus is a major risk factor for cardiovascular disease, especially coronary artery disease. This study aimed to evaluate the association between TCF7L2 polymorphism rs7903146 and coronary artery disease in diabetic and non-diabetic subjects. Methods and Results: two populations were studied in order to assess severity of coronary artery disease and cardiovascular events incidence. Eight-hundred and eighty nine subjects who were referred for cardiac catheterization for coronary artery disease diagnosis were cross-sectionally evaluated for coronary lesions (atherosclerotic burden) and 559 subjects from the MASS-II Trial were prospectively followed-up for 5 years and assessed for major cardiovascular events incidence. As expected, rs7903146 T allele was associated with diabetes. Although diabetic patients had a higher prevalence of coronary lesions, no association between TCF7L2 genotype and coronary lesions was found in this subgroup. However, non-diabetic individuals carrying the T allele were associated with a significantly higher frequency of coronary lesions than non-diabetic non-carriers of the risk allele (adjusted OR = 2.32 95% CI 1.27-4.24, p = 0.006). Moreover, presence of multi-vessel coronary artery disease was also associated with the CT or TT genotypes in non-diabetics. Similarly, from the prospective sample analysis, non-diabetics carrying the CT/TT genotypes had significantly more composite cardiovascular end-points events than CC carriers (p = 0.049), mainly due to an increased incidence of death (p = 0.004). Conclusions: rs7903146 T allele is associated with diabetes and, in non-diabetic individuals, with a higher prevalence and severity of coronary artery disease and cardiovascular events. name of registry site (see list below), registration number, trial registration URL in brackets.

Genetic Variants of Diabetes Risk and Incident Cardiovascular Events in Chronic Coronary Artery Disease

Objective: To determine whether information from genetic risk variants for diabetes is associated with cardiovascular events incidence. Methods: From the about 30 known genes associated with diabetes, we genotyped single-nucleotide polymorphisms at the 10 loci most associated with type-2 diabetes in 425 subjects from the MASS-II Study, a randomized study in patients with multi-vessel coronary artery disease. The combined genetic information was evaluated by number of risk alleles for diabetes. Performance of genetic models relative to major cardiovascular events incidence was analyzed through Kaplan-Meier curve comparison and Cox Hazard Models and the discriminatory ability of models was assessed for cardiovascular events by calculating the area under the ROC curve. Results: Genetic information was able to predict 5-year incidence of major cardiovascular events and overall-mortality in non-diabetic individuals, even after adjustment for potential confounders including fasting glycemia. Non-diabetic individuals with high genetic risk had a similar incidence of events then diabetic individuals (cumulative hazard of 33.0 versus 35.1% of diabetic subjects). The addition of combined genetic information to clinical predictors significantly improved the AUC for cardiovascular events incidence (AUC = 0.641 versus 0.610). Conclusions: Combined information of genetic variants for diabetes risk is associated to major cardiovascular events incidence, including overall mortality, in non-diabetic individuals with coronary artery disease.

Traffic-Related Air Pollution and Perinatal Mortality: A Case-Control Study

BACKGROUND: Ambient levels of air pollution may affect the health of children, as indicated by studies of infant and perinatal mortality. Scientific evidence has also correlated low birth weight and preterm birth, which are important determinants of perinatal death, with air pollution. However, most of these studies used ambient concentrations measured at monitoring sites, which may not consider differential exposure to pollutants found at elevated concentrations near heavy-traffic roadways. OBJECTIVES: Our goal was to examine the association between traffic-related pollution and perinatal mortality. METHODS: We used the information collected for a case-control study conducted in 14 districts in the City of Sao Paulo, Brazil, regarding risk factors for perinatal deaths. We geocoded the residential addresses of cases (fetal and early neonatal deaths) and controls (children who survived the 28th day of life) and calculated a distance-weighted traffic density (DWTD) measure considering all roads contained in a buffer surrounding these homes. RESULTS: Logistic regression revealed a gradient of increasing risk of early neonatal death with higher exposure to traffic-related air pollution. Mothers exposed to the highest quartile of the DWTD compared with those less exposed exhibited approximately 50% increased risk (adjusted odds ratio = 1.47; 95% confidence interval, 0.67-3.19). Associations for fetal mortality were less consistent. CONCLUSIONS: These results suggest that motor vehicle exhaust exposures may be a risk factor for perinatal mortality.

How much do smoking and alcohol consumption explain socioeconomic inequalities in head and neck cancer risk?

Background A higher burden of head and neck cancer has been reported to affect deprived populations. This study assessed the association between socioeconomic status and head and neck cancer, aiming to explore how this association is related to differences of tobacco and alcohol consumption across socioeconomic strata. Methods We conducted a case-control study in Sao Paulo, Brazil (1998-2006), including 1017 incident cases of oral, pharyngeal and laryngeal cancer, and 951 sex- and age-matched controls. Education and occupation were distal determinants in the hierarchical approach; cumulative exposure to tobacco and alcohol were proximal risk factors. Outcomes of the hierarchical model were compared with fully adjusted ORs. Results Individuals with lower education (OR 2.27; 95% CI 1.61 to 3.19) and those performing manual labour (OR 1.55; 95% CI 1.26 to 1.92) had a higher risk of disease. However, 54% of the association with lower education and 45% of the association with manual labour were explained by proximal lifestyle exposures, and socioeconomic status remained significantly associated with disease when adjusted for smoking and alcohol consumption. Conclusions Socioeconomic differences in head and neck cancer are partially attributable to the distribution of tobacco smoking and alcohol consumption across socioeconomic strata. Additional mediating factors may explain the remaining variation of socioeconomic status on head and neck cancer.

Penicillin-resistant pneumococcus and risk of treatment failure in pneumonia

Objective: To determine whether the presence of in vitro penicillin-resistant Streptococcus pneumoniae increases the risk of clinical failure in children hospitalised with severe pneumonia and treated with penicillin/ampicillin. Design: Multicentre, prospective, observational study. Setting: 12 tertiary-care centres in three countries in Latin America. Patients: 240 children aged 3-59 months, hospitalised with severe pneumonia and known in vitro susceptibility of S pneumoniae. Intervention: Patients were treated with intravenous penicillin/ampicillin after collection of blood and, when possible, pleural fluid for culture. The minimal inhibitory concentration (MIC) test was used to determine penicillin susceptibility of the pneumococcal strains isolated. Children were continuously monitored until discharge. Main outcome measures: The primary outcome was treatment failure (using clinical criteria). Results: Overall treatment failure was 21%. After allowing for different potential confounders, there was no evidence of association between treatment failure and in vitro resistance of S pneumoniae to penicillin according to the Clinical Laboratory Standards Institute (CLSI)/National Committee for Clinical Laboratory Standards (NCCLS) interpretative standards ((adj)RR = 1.03; 95%Cl: 0.49-1.90 for resistant S pneumoniae). Conclusions: Intravenous penicillin/ampicillin remains the drug of choice for treating penicillin-resistant pneumococcal pneumonia in areas where the MIC does not exceed 2 mu g/ml.

Trends in dermatomyositis- and polymyositis-related mortality in the state of Sao Paulo, Brazil, 1985-2007: multiple cause-of-death analysis

Background: Dermatomyositis (DM) and polymyositis (PM) are rare systemic autoimmune rheumatic diseases with high fatality rates. There have been few population-based mortality studies of dermatomyositis and polymyositis in the world, and none have been conducted in Brazil. The objective of the present study was to employ multiple-cause of-death methodology in the analysis of trends in mortality related to dermatomyositis and polymyositis in the state of Sao Paulo, Brazil, between 1985 and 2007. Methods: We analyzed mortality data from the Sao Paulo State Data Analysis System, selecting all death certificates on which DM or PM was listed as a cause of death. The variables sex, age and underlying, associated or total mentions of causes of death were studied using mortality rates, proportions and historical trends. Statistical analysis were performed by chi-square and H Kruskal-Wallis tests, variance analysis and linear regression. A p value less than 0.05 was regarded as significant. Results: Over a 23-year period, there were 318 DM-related deaths and 316 PM-related deaths. Overall, DM/PM was designated as an underlying cause in 55.2% and as an associated cause in 44.8%; among 634 total deaths females accounted for 71.5%. During the study period, age-and gender-adjusted DM mortality rates did not change significantly, although PM as an underlying cause and total mentions of PM trended lower (p < 0.05). The mean ages at death were 47.76 +/- 20.81 years for DM and 54.24 +/- 17.94 years for PM (p = 0.0003). For DM/PM, respectively, as underlying causes, the principal associated causes of death were as follows: pneumonia (in 43.8%/33.5%); respiratory failure (in 34.4%/32.3%); interstitial pulmonary diseases and other pulmonary conditions (in 28.9%/17.6%); and septicemia (in 22.8%/15.9%). For DM/PM, respectively, as associated causes, the following were the principal underlying causes of death: respiratory disorders (in 28.3%/26.0%); circulatory disorders (in 17.4%/20.5%); neoplasms (in 16.7%/13.7%); infectious and parasitic diseases (in 11.6%/9.6%); and gastrointestinal disorders (in 8.0%/4.8%). Of the 318 DM-related deaths, 36 involved neoplasms, compared with 20 of the 316 PM-related deaths (p = 0.03). Conclusions: Our study using multiple cause of deaths found that DM/PM were identified as the underlying cause of death in only 55.2% of the deaths, indicating that both diseases were underestimated in the primary mortality statistics. We observed a predominance of deaths in women and in older individuals, as well as a trend toward stability in the mortality rates. We have confirmed that the risk of death is greater when either disease is accompanied by neoplasm, albeit to lesser degree in individuals with PM. The investigation of the underlying and associated causes of death related to DM/PM broaden the knowledge of the natural history of both diseases and could help integrate mortality data for use in the evaluation of control measures for DM/PM.

Late Entry into HIV Care: Estimated Impact on AIDS Mortality Rates in Brazil, 2003-2006

Background: Worldwide, a high proportion of HIV-infected individuals enter into HIV care late. Here, our objective was to estimate the impact that late entry into HIV care has had on AIDS mortality rates in Brazil. Methodology/Principal Findings: We analyzed data from information systems regarding HIV-infected adults who sought treatment at public health care facilities in Brazil from 2003 to 2006. We initially estimated the prevalence of late entry into HIV care, as well as the probability of death in the first 12 months, the percentage of the risk of death attributable to late entry, and the number of avoidable deaths. We subsequently adjusted the annual AIDS mortality rate by excluding such deaths. Of the 115,369 patients evaluated, 50,358 (43.6%) had entered HIV care late, and 18,002 died in the first 12 months, representing a 16.5% probability of death in the first 12 months (95% CI: 16.3-16.7). By comparing patients who entered HIV care late with those who gained timely access, we found that the risk ratio for death was 49.5 (95% CI: 45.1-54.2). The percentage of the risk of death attributable to late entry was 95.5%, translating to 17,189 potentially avoidable deaths. Averting those deaths would have lowered the 2003-2006 AIDS mortality rate by 39.5%. Including asymptomatic patients with CD4(+) T cell counts >200 and <= 350 cells/mm(3) in the group who entered HIV care late increased this proportion by 1.8%. Conclusions/Significance: In Brazil, antiretroviral drugs reduced AIDS mortality by 43%. Timely entry would reduce that rate by a similar proportion, as well as resulting in a 45.2% increase in the effectiveness of the program for HIV care. The World Health Organization recommendation that asymptomatic patients with CD4(+) T cell counts <= 350 cells/mm(3) be treated would not have a significant impact on this scenario.

Risk factors for neonatal transfers from the Sapopemba free-standing birth centre to a hospital in Sao Paulo, Brazil

Objective: to identify risk factors associated with neonatal transfers from a free-standing birth centre to a hospital. Design: epidemiological case-control study. Setting: midwifery-led free-standing birth centre in Sao Paulo, Brazil. Participants: 96 newborns were selected from 2840 births between September 1998 and August 2005. Cases were defined as all new borns transferred from the birth centre to a hospital (n = 32), and controls were defined as new borns delivered at the same birth centre, during the same time period, and who had not been transferred to a hospital (n = 64). Measurements and findings: data were collected from medical records available at the birth centre. Univariate and multivariate analyses were performed using logistic regression. The multivariate analysis included outcomes with p<0.25, specifically: smoking during pregnancy, prenatal care appointments, labour complications, weight in relation to gestational age, and one-minute Apgar score. Of the foregoing outcomes, those that remained in the full regression model as a risk factor associated with neonatal transfer were: smoking during pregnancy [p = 0.009, odds ratio (OR) = 4.1,95% confidence interval (CI) 1.03-16.33], labour complications (p<0.001, OR = 5.5, 95% CI 1.06-28.26) and one-minute Apgar score <= 7 (p<0.001, OR = 7.8,95% CI 1.62-37.03). Key conclusions and implications for practice: smoking during pregnancy, labour complications and one-minute Apgar score <= 7 were confirmed as risk factors for neonatal transfer from the birth centre to a hospital. The identified risk factors can help to improve institutional protocols and formulate hypotheses for other studies. (C) 2009 Elsevier Ltd. All rights reserved.

Changes in estimated coronary risk in the 1980s: data from 38 populations in the WHO MONICA Project

The World Health Organization (WHO) MONICA Project is a 10-year study monitoring trends and determinants of cardiovascular disease in geographically defined populations. Data were collected from over 100 000 randomly selected participants in two risk factor surveys conducted approximately 5 years apart in 38 populations using standardized protocols. The net effects of changes in the risk factor levels were estimated using risk scores derived from longitudinal studies in the Nordic countries. The prevalence of cigarette smoking decreased among men in most populations, but the trends for women varied. The prevalence of hypertension declined in two-thirds of the populations. Changes in the prevalence of raised total cholesterol were small but highly correlated between the genders (r = 0.8). The prevalence of obesity increased in three-quarters of the populations for men and in more than half of the populations for women. In almost half of the populations there were statistically significant declines in the estimated coronary risk for both men and women, although for Beijing the risk score increased significantly for both genders. The net effect of the changes in the risk factor levels in the 1980s in most of the study populations of the WHO MONICA Project is that the rates of coronary disease are predicted to decline in the 1990s.

Refining the risk-benefit equation for thrombolysis: How to identity the low risk patient before administering thrombolytic therapy

In view of the relative risk of intracranial haemorrhage and major bleeding with thrombolytic therapy, it is important ro identify as early as possible the low risk patient who may not have a net clinical benefit from thrombolysis in the setting of acute myocardial infarction. An analysis of 5434 hospital-treated patients with myocardial infarction in the Perth MONICA study showed that age below 60 and absence of previous infarction or diabetes, shock, pulmonary oedema, cardiac arrest and Q-wave or left bundle branch block on the initial ECG identified a large group of patients with a 28 day mortality of only 1%, and one year mortality of only 2%. Identification of baseline risk in this way helps refine the risk-benefit equation for thrombolytic therapy, and may help avoid unnecessary use of thrombolysis in those unlikely to benefit.

Long-term risk of first recurrent stroke in the Perth Community Stroke Study

Background and Purpose-Few community-based studies have examined the long-term risk of recurrent stroke after an acute first-ever stroke. This study aimed to determine the absolute and relative risks of a first recurrent stroke over the first 5 years after a first-ever stroke and the predictors of such recurrence in a population-based series of people with first-ever stroke in Perth, Western Australia. Methods-Between February 1989 and August 1990, all people with a suspected acute stroke or transient ischemic attack of the brain who were resident in a geographically defined region of Perth, Western Australia, with a population of 138 708 people, were registered prospectively and assessed according to standardized diagnostic criteria. Patients were followed up prospectively at 4 months, 12 months, and 5 years after the index event. Results-Three hundred seventy patients with a first-ever stroke were registered, of whom 351 survived >2 days. Data were available for 98% of the cohort at 5 years, by which time 199 patients (58%) had died and 52 (15%) had experienced a recurrent stroke, 12 (23%) of which were fatal within 28 days. The 5-year cumulative risk of first recurrent stroke was 22.5% (95% confidence limits [CL], 16.8%, 28.1%). The risk of recurrent stroke was greatest in the first 6 months after stroke, at 8.8% (95% CL, 5.4%, 12.1%). After adjustment for age and sex, the prognostic factors for recurrent stroke were advanced, but not extreme, age (75 to 84 years) (hazard ratio [HR], 2.6; 95% CL, 1.1, 6.2), hemorrhagic index stroke (HR, 2.1; 95% CL, 0.98, 4.4), and diabetes mellitus (HR, 2.1; 95% CL, 0.95, 4.4). Conclusions-Approximately 1 in 6 survivors (15%) of a first-ever stroke experience a recurrent stroke over the next 5 years, of which 25% are fatal within 28 days. The pathological subtype of the recurrent stroke is the same as that of the index stroke in 88% of cases. The predictors of first recurrent stroke in this study were advanced age, hemorrhagic index stroke, and diabetes mellitus, but numbers of recurrent events were modest. Because the risk of recurrent stroke is highest (8.8%) in the first 6 months after stroke, strategies for secondary prevention should be initiated as soon as possible after the index event.