Back

Detecting early loss of kidney function in a Sri Lankan cohort study of working age adults

Chulasiri, P.; Rutter, C. E.; Gunawardena, N.; Wickramasinghe, S.; Niyas, R.; Pearce, N.; Caplin, B.; Ruwanpathirana, T.

2026-08-19 epidemiology
10.64898/2026.08.18.26360601 medRxiv
Show abstract

Background: Chronic kidney disease of undetermined cause (CKDu) is a form of kidney disease not associated with traditional risk factors such as hypertension, diabetes or heavy proteinuria. 11.2% of men and 3.7% of women demonstrated low eGFR (a surrogate for CKDu) in the absence of these risk factors in a 2017 cross-sectional population-representative survey of adults in North Central Province, Sri Lanka. We therefore established a longitudinal cohort to track changes in kidney function over time and to identify risk factors for developing poor kidney health. Methods: This was a 6-year study of adults aged 20-60 years conducted in Puhudivula, Anuradhapura district. Exclusions included evidence of diabetes, hypertension or pre-existing CKD. We fitted hidden Markov models (HMMs) to estimate underlying state of kidney health and examine risk factors associated with departure from a healthy state. Results: We identified four kidney health trajectories in the population (n=425): always healthy (74%); unhealthy throughout (5%); transition from health to unhealthy (10%); and reversion from unhealthy to healthy (11%). Using smokeless tobacco, including betel quid, was associated with being in an unhealthy category (2.29 [1.17, 4.49]). Lagged exposure to smoking (2.26 [1.25, 4.10]), smokeless tobacco (1.98 [1.13, 3.48]) and weedkiller (1.72 [1.15, 2.59]) were associated with the point of transition to an unhealthy state. Conclusions: Almost a quarter of working age adults in this population demonstrated eGFR changes consistent with poor kidney health. Smokeless tobacco use was associated with both pre-existing evidence of poor kidney health and transitioning to the unhealthy category.

Matching journals

The top 5 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.