Development and validation of a mobile application based on a machine learning model to aid in predicting dosage of vitamin K antagonists among Indian patients post mechanical heart valve replacement

Patients who undergo heart valve replacements with mechanical valves need to take Vitamin K Antagonists (VKA) drugs (Warfarin, Nicoumalone) which has got a very narrow therapeutic range and needs very close monitoring using PT-INR. Accessibility to physicians to titrate drugs doses is a major problem in low-middle income countries (LMIC) like India. Our work was aimed at predicting the maintenance dosage of these drugs, using the de-identified medical data collected from patients attending an INR Clinic in South India. We used artificial intelligence (AI) - machine learning to develop the algorithm. A Support Vector Machine (SVM) regression model was built to predict the maintenance dosage of warfarin, who have stable INR values between 2.0 and 4.0. We developed a simple user friendly android mobile application for patients to use the algorithm to predict the doses. The algorithm generated drug doses in 1100 patients were compared to cardiologist prescribed doses and found to have an excellent correlation.

Prevalence of metabolic syndrome and its risk factors in Kerala, South India: Analysis of a community based cross-sectional study

BACKGROUND

Coronary Artery Disease (CAD) is a leading cause of death and disability in Kerala, India. Metabolic syndrome (MS) is a constellation of established risk factors for CAD. We aimed to estimate the prevalence of MS and evaluate the association between MS and CAD using a community-based sample population.

METHODS

A cross-sectional community based survey was conducted in urban and rural areas of Kerala in 2011. We included 5063 individuals for analysis. Age standardized prevalence of MS, associated diagnoses (hypertension, diabetes and hypercholesterolemia) and other potential risk factors were assessed for men and women in both urban and rural locations. Univariate and multivariate logistic regression models were developed to identify participant characteristics that are associated with MS.

RESULTS

After standardization for age and adjustment for sex and urban-rural distribution, the prevalence of metabolic syndrome in Kerala was 24%, 29% and 33% for the NCEP ATP III, IDF and AHA/NHLBI Harmonization definitions, respectively. The mean (SD) age of the participants was 51 (14) years, and 60% were women. Women had a higher prevalence of MS than men (28% versus 20% for ATP III, p<0.001). Similarly, participants living in urban areas had higher prevalence of MS than their rural counterparts (26% versus 22%, p<0.001). Elevated body mass index, older age, and female sex were associated with MS in an adjusted multivariate model. The propensity for definite CAD was 1.7 times higher in individuals with MS defined based on ATP III criteria compared to those without MS (Adjusted OR = 1.69; 95% CI: 1.3-2.2, p<0.001).

CONCLUSIONS

One of four to one of three adult individuals in Kerala have MS based on different criteria. Higher propensity for CAD in individuals with MS in Kerala calls for urgent steps to prevent and control the burden of metabolic conditions.

Diffuse reflectance spectroscopy: diagnostic accuracy of a non-invasive screening technique for early detection of malignant changes in the oral cavity

BACKGROUND

Strong proof-of-principle for utilisation of diffuse reflectance spectroscopy, a non-invasive tool for early detection of malignant changes, has emerged recently. The potential of this technique in distinguishing normal tissue from hyperplastic and dysplastic tissues was explored.

METHODS

Diffuse reflectance (DR) spectra in the 400-700 nm region were obtained from the buccal mucosa of 96 patients and 34 healthy volunteers. The DR spectral data were compared against the gold standard biopsy and histopathology results. A principal-component analysis was performed for dimensional reduction in the normalised spectral data with linear discriminant analysis as the classifying technique. The receiver operator characteristic curve technique was employed for evaluating the performance of the diagnostic test.

RESULTS

DR spectral features for different lesions, such as normal/healthy, hyperplastic, dysplastic and squamous cell carcinoma (SCC), varied significantly according to the intensity of oxygenated haemoglobin absorption. While the classification based on discriminant scores provided an overall sensitivity of 98.5% and specificity of 96.0% for distinguishing SCC from dysplasia, they were 100.0% and 95.0%, respectively, for distinguishing dysplasia from hyperplasia. Similarly, the analysis yielded a sensitivity of 95.0% and specificity of 100.0% for distinguishing hyperplasia from healthy tissue. The areas under the receiver operator characteristic curves were 0.98 (95% CI 0.95 to 1.00) and 0.95 (95% CI 0.90 to 1.00) for distinguishing dysplasia from SCC and hyperplasia from dysplasia, respectively.

CONCLUSION

DR spectral data efficiently discriminate healthy tissue from oral malignant lesions. Diagnostic accuracies obtained in this study highlight the potential use of this method for routine clinical practice.

Validation of self-reported smoking status by measuring serum cotinine levels: an Indian perspective

BACKGROUND

Serum cotinine levels are a reliable marker of tobacco use. Few studies have validated questionnaires assessing smoking and exposure to environmental tobacco smoke (ETS) against serum levels. We undertook such a study in industrial workers in India.

METHODS

We chose 426 individuals by stratified random sampling from a database of 3397 individuals surveyed at New Delhi for the cardiovascular disease surveillance programme in a large industrial setting. Questionnaires assessing details of smoking practices and duration of exposure to ETS (if any) were administered. Cotinine levels were measured in the blood samples of these individuals.

RESULTS

The study population comprised 142 nonsmokers not exposed to ETS, 142 non-smokers exposed to ETS and 142 active smokers. Cotinine levels among nonsmokers not exposed to ETS were non-detectable; and for non-smokers exposed to ETS and active smokers, the median (interquartile range) levels were non-detectable (non-detectable to 46.1 ng/ml) and 336 ng/ml (204-500 ng/ml), respectively. The best combined sensitivity (91%) and specificity (87.2%) yielded a cotinine cut-off level of 40.35 ng/ml to differentiate active smokers from non-smokers not exposed to ETS and those exposed to ETS (area under the curve 0.902). The cut-off cotinine level was estimated at 10.95 ng/ml using a similar analysis (sensitivity 43%, specificity 82%; area under the curve 0.64) to distinguish non-smokers not exposed to ETS from those exposed to ETS. The misclassification rate was estimated at 19% and 57.1% among self-reported non-smokers not exposed to ETS and those exposed to ETS, respectively.

CONCLUSIONS

Obtaining a history of tobacco use is an accurate method of detecting smokers in epidemiological studies whereas serum cotinine levels accurately differentiate smokers from non-smokers. However, a brief questionnaire assessing passive exposure to smoke has poor sensitivity in distinguishing non-smokers exposed to ETS from those not exposed to ETS.

Impact of alcohol on coronary heart disease in Indian men

BACKGROUND

Moderate alcohol consumption is known to be protective against coronary heart disease (CHD). However, the INTERHEART study, a case-control study of acute myocardial infarction (MI) patients, revealed that alcohol consumption in South Asians was not protective against CHD. We therefore planned to study cardiovascular risk factor and CHD prevalence among male alcohol users as compared to age matched lifetime abstainers.

METHODS

The subjects for this study were recruited from a cross-sectional survey carried out among employees and their family members aged 20-69 years in 10 medium-to-large industries from diverse sites in India, using a stratified random sampling technique. Information on education, behavioral, clinical and biochemical risk factors of CHD and alcohol use was obtained through standardized instruments. CHD diagnosis was based on Rose Questionnaire or a prior physician diagnosed CHD.

RESULTS

A total of 4465 subjects were present or past alcohol users. The mean age of alcohol users and lifetime abstainers was 42.8+/-11.0 years and 42.8+/-11.1 years, respectively (p=0.90). Systolic blood pressure and diastolic blood pressure were significantly higher in alcohol users (128.7+/-17.6 mmHg/80.1+/-11.3 mmHg) as compared to lifetime abstainers (126.9+/-15.9 mmHg/79.5+/-10.3 mmHg, p<0.01). Fasting blood sugar in alcohol users (98.7+/-30.5 mg%) was also significantly higher than lifetime abstainers (96.6+/-26.0 mg%, p<0.01). Total cholesterol was lower in alcohol users (179.1+/-41.1 mg%) as compared to lifetime abstainers (182.7+/-38.2 mg%, p<0.01). HDL cholesterol was higher in alcohol users (42.9+/-10.8 mg%) as compared to lifetime abstainers (41.3+/-10.0 mg%, p<0.01). Body mass index (BMI) was lower in alcohol users as compared to lifetime abstainers (22.7+/-4.1 kg/m2 vs. 24.0+/-3.3 kg/m2, p<0.001). Tobacco use was significantly higher in alcohol users (63.1% vs. 20.7%). The odds ratio (OR) of having CHD after adjusting for tobacco use, BMI and education was 1.4 (95%CI 1.0-1.9) in alcohol users as compared to controls. The OR was 1.2 (95%CI 0.8-1.6) in occasional alcohol users, 1.6 (95%CI 1.0-2.2) in regular alcohol users and 2.1 (95% CI 1.1-3.0) in past alcohol users as compared to controls.

CONCLUSION

We did not observe an inverse (protective) association between alcohol intake and the prevalence of CHD. In contrast, our study indicated an association in the reverse direction, suggesting possible harm of alcohol for coronary risk in Indian men. This relationship needs to be further examined in large, prospective study.

Double burden of underweight and overweight among children (10-19 years of age) of employees working in Indian industrial units

BACKGROUND

Along with the existing problem of underweight, overweight in children is increasing in the developing world. However, there is little information on its magnitude and pattern in the Indian context. We aimed to study the pattern and correlates of overweight in Indian children and adolescents.

METHODS

A total of 3750 children in the age group of 10-19 years, who were family members of randomly selected employees from 10 different industrial sites in India, were surveyed using an interviewer-administered questionnaire.

RESULTS

The prevalence of underweight was highest in peri-urban areas (30.2% and 53.2% according to Indian and international criteria, respectively). In urban and highly urban areas, the prevalence of underweight was 14.1% and 9.8%, respectively, according to the Indian criteria, and 27.1% and 19.2%, respectively, according to international criteria. The proportion of overweight children was highest in the highly urban category (19.1% and 13.4% according to Indian and international criteria, respectively). The level of urbanization (OR 3.1 and 4.7 for overweight in urban and highly urban areas, respectively, compared with peri-urban areas, p < 0.001), physical activity (OR 0.4, p < 0.001, in children with physical activity score > or = 75th percentile compared with a score < or = 75th percentile) and frequency of meals outside the home (OR 12, p < 0.001, if > 25% weekly meals taken outside the home compared with < 25% of weekly meals outside home) were significant predictors of overweight.

CONCLUSION

There is a double burden of underweight and overweight among Indian children and adolescents.

Prevalence and determinants of diabetes mellitus in the Indian industrial population

AIM

To highlight the regional difference in the prevalence of diabetes mellitus (DM) and to explore determinants in variability in the Indian industrial population.

METHODS

A cross-sectional survey was carried out among the employees and their family members (10 930 individuals, mean age 39.6 years, 6764 male) of eleven medium-to-large industries from diverse sites in India, using a stratified random sampling technique. Information on behavioural, clinical and biochemical risk factors of DM was obtained, through standardized instruments. DM was diagnosed when fasting blood glucose was > or = 7.0 mmol/l and/or individuals took drug treatment for DM. Multiple logistic regression analysis was carried out to identify the potential predictors of DM.

RESULT

In the 20 to 69-year-old age group, the crude prevalence of DM and impaired fasting glucose was 10.1 and 5.3%, respectively. Urban sites had a higher prevalence and awareness of DM status. Individuals in the lower education group had a high prevalence of DM (11.6%). In diabetic subjects, 38.4% were unaware that they had diabetes. Waist-circumference-to-height ratio had a higher DM predictive power than waist circumference and body mass index. The risk factors associated with overall prevalence of DM were: age, sex, low-education level, family history of DM, hypertension and overweight/obesity. Interaction of risk factors was observed only in urban high-prevalence sites.

CONCLUSION

There are wide regional variations in the prevalence of DM in India. The high burden of undetected diabetes, even in settings with universal access to on-site health care, highlights the need for innovative prevention and control strategies.

Management of acute coronary syndromes in secondary care settings in Kerala: impact of a quality improvement programme

BACKGROUND

Evidence-based therapies that have been shown to improve outcomes in acute coronary syndromes (ACS) are often underused in clinically eligible patients. We evaluated the impact, efficacy and acceptability of a quality improvement programme to manage ACS.

METHODS

A well-defined geographical area was identified and a situational analysis done. All physicians in the area, who were actively involved in the detection and management of ACS, were invited to participate in the quality improvement programme. The programme involved the use of a service delivery package which consisted of standard admission orders and patient-directed discharge instructions. Concurrently, health education in the community to promote self-detection, self-administration of aspirin and self-referral were carried out. All participating physicians were asked to register consecutive cases of ACS (20 each) presenting to their clinics before and after the intervention programme. The pre- and post-intervention data were compared.

RESULTS

The use of aspirin at discharge increased from 89.7% to 96.8% (p < 0.05) and that of heparin from 57.6% to 66.3% (p < 0.05). The use of beta-blockers increased from 48.6% to 63.4% (p < 0.05) and that of lipid-lowering therapy from 74.1% to 96.3% (p < 0.05). There was a significant reduction in the use of calcium channel blockers from 21.6% to 8.1% (p < 0.05). The time to thrombolysis decreased significantly (median difference of 54 minutes, p < 0.05) after the intervention programme.

CONCLUSION

Structured quality improvement programmes aimed at both patients and providers can be successful in secondary care settings of developing countries.

Risk factors related to the development of diabetes in men working in a north Indian industry

BACKGROUND

Epidemiological and lifestyle changes have been implicated in the high burden of diabetes in urban India. However, longitudinal data on the determinants for the development of diabetes in this population are not available. We investigated the determinants for the development of diabetes in workers in an Indian industrial organization.

METHODS

Two cross-sectional surveys were done, using similar methodology (Survey 1 during 1995-98 [n=2548] and Survey 2 during 2002-03 [n=2800]) among all employees (age 20-59 years) of an industrial organization. A large majority of these were men (89.5% in Survey 1 and 92.8% in Survey 2). Men with no diabetes at baseline, who participated in both the surveys (n=942), constituted the study population. Development of new-onset diabetes was defined using history and fasting glucose concentrations > or =7 mmol/L.

RESULTS

The mean (SD) age of the participants at baseline was 40 (2) years. Diabetes developed in 8% of the study population over 6.8 (1.7) years. Individuals who developed diabetes had significantly higher age, blood pressure, body mass index, waist circumference, fasting and post-prandial glucose, post-prandial insulin and fasting triglyceride levels at baseline. On multivariate regression analysis, only impaired glucose tolerance (OR 3.8, 95% CI: 2.1-6.8) and waist circumference (OR 1.09, 95% CI: 1.02-1.16) predicted the development of diabetes. Presence of the metabolic syndrome, as defined by the modified National Cholesterol Education Program Adult Treatment Panel (NCEP-ATP) III and WHO criteria, increased the odds (95% CI) of developing diabetes by 2.2 (1.3-3.6) and 4.5 (2.7-7.4) times, respectively.

CONCLUSION

Impaired glucose tolerance, high waist circumference and the metabolic syndrome are powerful predictors for the development of diabetes among urban Indian men.

Methods for establishing a surveillance system for cardiovascular diseases in Indian industrial populations

OBJECTIVE

To establish a surveillance network for cardiovascular diseases (CVD) risk factors in industrial settings and estimate the risk factor burden using standardized tools.

METHODS

We conducted a baseline cross-sectional survey (as part of a CVD surveillance programme) of industrial populations from 10 companies across India, situated in close proximity to medical colleges that served as study centres. The study subjects were employees (selected by age and sex stratified random sampling) and their family members. Information on behavioural, clinical and biochemical determinants was obtained through standardized methods (questionnaires, clinical measurements and biochemical analysis). Data collation and analyses were done at the national coordinating centre.

FINDINGS

We report the prevalence of CVD risk factors among individuals aged 20-69 years (n = 19 973 for the questionnaire survey, n = 10 442 for biochemical investigations); mean age was 40 years. The overall prevalence of most risk factors was high, with 50.9% of men and 51.9% of women being overweight, central obesity was observed among 30.9% of men and 32.8% of women, and 40.2% of men and 14.9% of women reported current tobacco use. Self-reported prevalence of diabetes (5.3%) and hypertension (10.9%) was lower than when measured clinically and biochemically (10.1% and 27.7%, respectively). There was marked heterogeneity in the prevalence of risk factors among the study centres.

CONCLUSION

There is a high burden of CVD risk factors among industrial populations across India. The surveillance system can be used as a model for replication in India as well as other developing countries.