Futurists have anticipated that novel autonomous technologies, embedded with machine learning (ML), will substantially influence healthcare. ML is focused on making predictions as accurate as possible, while traditional statistical models are aimed at inferring relationships between variables. The benefits of ML comprise flexibility and scalability compared with conventional statistical approaches, which makes it deployable for several tasks, such as diagnosis and classification, and survival predictions. However, much of ML-based analysis remains scattered, lacking a cohesive structure. There is a need to evaluate and compare the performance of well-developed conventional statistical methods and ML on patient outcomes, such as survival, response to treatment, and patient-reported outcomes (PROs). In this article, we compare the usefulness and limitations of traditional statistical methods and ML, when applied to the medical field. Traditional statistical methods seem to be more useful when the number of cases largely exceeds the number of variables under study and a priori knowledge on the topic under study is substantial such as in public health. ML could be more suited in highly innovative fields with a huge bulk of data, such as omics, radiodiagnostics, drug development, and personalized treatment. Integration of the two approaches should be preferred over a unidirectional choice of either approach.

OBJECTIVE: This population-based study, carried out in the framework of the Verona Diabetes Study, investigated mortality from specific causes in known type 2 diabetic patients. RESEARCH DESIGN AND METHODS: A cohort of 7,148 known type 2 diabetic patients (3,366 men and 3,782 women) was identified on 31 December 1986 and followed up for 5 years (1987-1991). Underlying causes of death were obtained from death certificates and were coded according to the International Classification of Diseases, Ninth Revision. Cause-specific death rates of diabetic subjects were compared with those of the inhabitants of Verona. By 31 December 1991, 1,550 diabetic subjects (744 men and 806 women) had died. RESULTS: The standardized mortality ratio (SMR) for all causes of death was 1.42 (95% CI 1.35-1.50). The highest SMRs were for the following specific causes: diabetes (SMR 4.47 [3.91-5.10]), gastrointestinal diseases (1.83 [1.50-2.21])--particularly liver cirrhosis (2.52 [1.96-3.20])--and cardiovascular diseases (1.34 [1.23-1.44]), particularly cerebrovascular (1.48 [1.25-1.73]) and ischemic heart diseases (1.41 [1.24-1.62]). A significantly higher than expected risk of mortality for cardiovascular causes was already present in the first 5 years after diagnosis and decreased with age. Type 2 diabetic patients treated with insulin had a higher risk of dying than those treated orally or by diet. CONCLUSIONS: The highest SMRs in the diabetic cohort were for diabetes and liver cirrhosis. The mortality risk for cardiovascular diseases, although significantly higher than expected, was much lower in Italian type 2 diabetic patients than that reported for American patients. The evidence of an early effect on mortality suggests that prevention, early diagnosis, and treatment should be improved.

The joint distribution of asthma and chronic obstructive pulmonary disease (COPD) has not been well described. This study aims at determining the prevalence of self-reported physician diagnoses of asthma, COPD and of the asthma-COPD overlap syndrome and to assess whether these conditions share a common set of risk factors. A screening questionnaire on respiratory symptoms, diagnoses and risk factors was administered by mail or phone to random samples of the general Italian population aged 20–44 (n = 5163) 45–64 (n = 2167) and 65–84 (n = 1030) in the frame of the multicentre Gene Environment Interactions in Respiratory Diseases (GEIRD) study. A physician diagnosis of asthma or COPD (emphysema/chronic bronchitis/COPD) was reported by 13% and 21% of subjects aged <65 and 65–84 years respectively. Aging was associated with a marked decrease in the prevalence of diagnosed asthma (from 8.2% to 1.6%) and with a marked increase in the prevalence of diagnosed COPD (from 3.3% to 13.3%). The prevalence of the overlap of asthma and COPD was 1.6% (1.3%–2.0%), 2.1% (1.5%–2.8%) and 4.5% (3.2%–5.9%) in the 20–44, 45–64 and 65–84 age groups. Subjects with both asthma and COPD diagnoses were more likely to have respiratory symptoms, physical impairment, and to report hospital admissions compared to asthma or COPD alone (p<0.01). Age, sex, education and smoking showed different and sometimes opposite associations with the three conditions. Asthma and COPD are common in the general population, and they coexist in a substantial proportion of subjects. The asthma-COPD overlap syndrome represents an important clinical phenotype that deserves more medical attention and further research.

OBJECTIVE: In the present study, we evaluated whether the coefficient of variation (CV) of fasting plasma glucose (FPG) over a 3-year period was a significant predictor of mortality in type 2 diabetic patients aged 56-74 years. RESEARCH DESIGN AND METHODS: All type 2 diabetic patients (n = 1,409) aged 56-74 years attending the Verona Diabetes Clinic and having at least two FPG determinations in each of the years 1984-1986 were followed for 10 years (1987-1996) to assess total and cause-specific mortality Patients were grouped into tertiles of mean and CV of FPG during 1984-1986. These parameters as well as sex, age, diabetes duration, insulin treatment, smoking, hypertension, and hypercholesterolemia were included in multivariate survival analyses. RESULTS: During the follow-up, 468 patients died. The CV of FPG was an independent predictor of total, cardiovascular, and cancer mortality. Mean FPG was a predictor of total mortality only when the CV of FPG was not included in the analyses. CONCLUSIONS: Long-term variability of fasting glucose is an independent predictor of mortality in patients with type 2 diabetes. The CV of FPG might be considered a useful additional parameter in the management of these patients.

The prevalence of asthma increased worldwide until the 1990s, but since then there has been no clear temporal pattern.The present study aimed to assess time trends in the prevalence of current asthma, asthma-like symptoms and allergic rhinitis in Italian adults from 1990 to 2010.The same screening questionnaire was administered by mail or phone to random samples of the general population (age 20–44 yrs) in Italy, in the frame of three multicentre studies: the European Community Respiratory Health Survey (ECRHS) (1991–1993; n=6,031); the Italian Study on Asthma in Young Adults (ISAYA) (1998–2000; n=18,873); and the Gene Environment Interactions in Respiratory Diseases (GEIRD) study (2007–2010; n=10,494). Time trends in prevalence were estimated using Poisson regression models in the centres that repeated the survey at different points in time.From 1991 to 2010, the median prevalence of current asthma, wheezing and allergic rhinitis increased from 4.1% to 6.6%, from 10.1% to 13.9% and from 16.8% to 25.8%, respectively. The prevalence of current asthma was stable during the 1990s and increased (relative risk 1.38, 95% CI 1.19–1.59) from 1998–2000 to 2007–2010, mainly in subjects who did not report allergic rhinitis. The prevalence of allergic rhinitis has increased continuously since 1991.The asthma epidemic is not over in Italy. During the past 20 yrs, asthma prevalence has increased by 38%, in parallel with a similar increase in asthma-like symptoms and allergic rhinitis.