Overview

My current research takes a cross-disciplinary approach to tackle problems, and focuses on the process of extracting useful and actionable information from data to aid various stakeholders in public health to make informed decisions. My main contribution to public health research is to contextualize the research problems under an analytical framework that facilitates empirical research. Thus allowing us to apply quantitative methods to data from different sources (e.g., registry data, cohort data, electronic health records, wearable device) and transform data into knowledge to improve human lives. As such, the spectrum of my applied research activities is broad (e.g., chronic disease epidemiology; health promotion in physical activity and cognition among the elderly). Such diverse research activities have a positive influence on my methods research in biostatistics, data analytics and health informatics. My main contributions in this area include the development of seamless and robust statistical workflows for observational studies with continuous or ordinal outcomes, the development of techniques that avoid dichotomization or transformation of continuous outcomes in the data analysis step, the development of software programs, and the novel application of quantitative and visualization approaches.

Research objectives

• To advance knowledge that translates into actionable insights for improving human health through transdisciplinary research.
• To apply and develop quantitative and data-centric methodologies and tools.

Publications 

List of publications in pdf file

Research highlight 1: Develop seamless and robust analytical workflow for studies with continuous or ordinal outcomes
The proposed analytical workflow focuses on assessing the existence of an exposure effect in its first step. It leverages on stratification for confounder adjustment, thereby modeling the ordering of the outcomes within each stratum. The likelihood of the ordered outcomes is equivalent to a discrete-choice model (i.e., rank-ordered logit model) and it returns an estimated scaled exposure effect from the maximum likelihood inference. The subsequent two steps of the workflow focus on quantifying the exposure effect on the original continuous scale by estimating the scale parameter of the error distribution (i.e., Extreme Value Type 1) and assessing the adequacy of the distributional assumption of the error terns with residual diagnostics. The proposed analytical workflow extends the benefits of matching for binary outcomes to continuous outcomes and subsequently ordinal outcomes by assuming that such outcomes are imperfectly-observed tied continuous outcome. Given that the likelihood of the rank-ordered logit model is a special case of a stratified Cox model, this body of work highlighted a common class of regression models is available for a wide range of outcomes (i.e., binary, survival time, continuous and ordinal) in stratified analysis. Recent extension of this work generalized the analytical workflow to normal error distribution with two repeated continuous measurements while allowing for any monotonic transformation on the outcomes via Box-Cox transformation.

• Tan CS, Støer NC, Chen Y, Andersson M, Ning Y, Wee HL, Khoo EYH, Tai ES, Kao SL, Reilly M. A stratification approach using logit-based models for confounder adjustment in the study of continuous outcomes. Stat Methods Med Res. 2019 Apr;28(4):1105-1125.
• Ning Y, Tan CS*, Maraki A, Ho PJ, Hodgins S, Comasco E, Nilsson KW, Wagner P, Khoo EY, Tai ES, Kao SL, Hartman M, Reilly M, Støer NC. Handling ties in continuous outcomes for confounder adjustment with rank-ordered logit and its application to ordinal outcomes. Stat Methods Med Res. 2020 Feb;29(2):437-454. *: Joint corresponding author.
• Ning Y, Støer NC, Ho PJ, Kao SL, Ngiam KY, Khoo EYH, Lee SC, Tai ES, Hartman M, Reilly M, Tan CS. Robust estimation of the effect of an exposure on the change in a continuous outcome. BMC Med Res Methodol. 2020 Jun 6;20(1):145.

Research highlight 2: Alleviate challenges in everyday, routine data analysis
When planning a data analysis, we often pose ourselves a series of probing questions. But the choice of our answer to each question will branch out to a different analytical strategy with its own advantages and disadvantages. In my applied research work, these are some analyses with multiple analytical alternatives available: (i) analyzing dichotomized outcome with several thresholds, (ii) quantifying temporal trends in age-standardized rates of diseases, and (iii) analyzing ordinal outcomes. For each analysis listed, I strive to strike a balance between theory and practice while facilitating the interpretation of findings.

• Tan CS, Støer N, Ning Y, Chen Y, Reilly M. Quantifying temporal trends of age-standardized rates with odds. Popul Health Metr. 2018 Dec 18;16(1):18..
• Chen Y, Ning Y, Kao SL, Støer NC, Müller-Riemenschneider F, Venkataraman K, Khoo EYH, Tai ES, Tan CS. Using marginal standardisation to estimate relative risk without dichotomising continuous outcomes. BMC Med Res Methodol. 2019 Jul 29;19(1):165.
• Ning Y, Ho PJ, Støer NC, Lim KK, Wee HL, Hartman M, Reilly M, Tan CS. A new procedure to assess when estimates from the cumulative link model can be interpreted as differences for ordinal scales in quality of life studies. Clin Epidemiol. 2021 Feb 4;13:53-65.

Research highlight 3: Analytical approaches for new data sources in public health research
With the tsunami of new data sources for public health research, this open up opportunities for us to extract invaluable information for various stakeholders in public health. Hence, I work with collaborators to apply and develop analytical methods for some of these new data sources.

• Chen Y, Kao SL, Tai ES, Wee HL, Khoo EY, Ning Y, Salloway MK, Deng X, Tan CS. Utilizing distributional analytics and electronic records to assess timeliness of inpatient blood glucose monitoring in non-critical care wards. BMC Med Res Methodol. 2016 Apr 8;16:40.
• Chen Y, Kao SL, Tan M, Ning Y, Salloway M, Wee HL, Venkataraman K, Khoo EYH, Chow YL, Tai ES, Tan CS. Feasibility of representing adherence to blood glucose monitoring through visualizations: A pilot survey study among healthcare workers. Int J Med Inform. 2018 Dec;120:172-178.
• Chen Y, Ning Y, Thomas P, Salloway M, Tan MLS, Tai ES, Kao SL, Tan CS. An open source tool to compute measures of inpatient glycemic control: translating from healthcare analytics research to clinical quality improvement. JAMIA Open. 2021 Jun 16;4(2):ooab033.

Research highlight 4: Applications of analytical approaches to chronic diseases and other conditions
Together with my collaborators, we conduct research in diabetes, stroke and cognition decline. For example, we investigated the inpatient diabetics program, and the role of caregiver and stroke factors on post-stroke’s healthcare utilization in Singapore.

• Tyagi S, Koh GC, Luo N, Tan KB, Hoenig H, Matchar DB, Yoong J, Chan A, Lee KE, Venketasubramanian N, Menon E, Chan KM, De Silva DA, Yap P, Tan BY, Chew E, Young SH, Ng YS, Tu TM, Ang YH, Kong KH, Singh R, Merchant RA, Chang HM, Yeo TT, Ning C, Cheong A, Ng YL, Tan CS. Role of caregiver factors in outpatient medical follow-up post-stroke: observational study in Singapore. BMC Fam Pract. 2021 Apr 14;22(1):74.
• Kao SL, Chen Y, Ning Y, Tan M, Salloway M, Khoo EYH, Tai ES, Tan CS. Evaluating the effectiveness of a multi-faceted inpatient diabetes management program among hospitalised patients with diabetes mellitus. Clin Diabetes Endocrinol. 2020 Nov 5;6(1):21.
• Hilal S, Tan CS#, Adams HHH, Habes M, Mok V, Venketasubramanian N, Hofer E, Ikram MK, Abrigo J, Vernooij MW, Chen C, Hosten N, Volzke H, Grabe HJ, Schmidt R, Ikram MA. Enlarged perivascular spaces and cognition: A meta-analysis of 5 population-based studies. Neurology. 2018 Aug 28;91(9):e832-e842. #: Joint first author