Background The impact of cold spells on population health can be categorized into an independent main effect of extreme low temperatures and an added effect of prolonged low temperatures. However, studies on the added effects of cold spells on hospitalizations remain limited.

Objective To investigate the added effects of cold spells on hospitalizations of residents in Hengyang City, Hunan Province, and to provide a scientific basis for establishing a cold spell early warning system.

Methods Daily meteorological data, air pollutant data, and hospitalization data from six tertiary hospitals of four districts in Hengyang City from 2017 to 2023 were collected. A generalized linear model (GLM) combined with a distributed lag nonlinear model (DLNM) was used to assess the added effects of cold spells on non-accidental hospitalizations, as well as hospitalizations for circulatory system diseases and respiratory system diseases, after controlling the main effect of temperature. The modifying effects of cold spell characteristics (intensity and duration) and individual characteristics (gender and age) were also analyzed.

Results Compared with non-cold spell periods, the relative risks (RRs) of total non-accidental hospitalizations and hospitalizations across disease categories, genders, and age groups were elevated during cold spells of varying intensities and durations. However, the total effects of cold spells exhibited a "U-shape" nonlinear relationship with intensity and decreased with prolonged duration. During high-intensity cold spells (daily average temperature < P₅ and lasting ≥ 2 d), the RR (95%CI) for non-accidental hospitalizations was 1.71 (1.21, 2.42); the RRs (95%CIs) for males and females were 1.99 (1.38, 2.84) and 1.47 (1.00, 2.16), respectively; for individuals < 65 years and ≥ 65 years, the RRs (95%CIs) were 1.59 (1.12, 2.26) and 1.93 (1.27, 2.92), respectively; and for circulatory and respiratory system diseases, the RRs (95%CIs) were 1.84 (1.22, 2.79) and 1.07 (0.71, 1.60), respectively. No statistically significant differences were observed between the above subgroups. The single-day lagged effects of cold spells displayed a two-peaked pattern. The single-day lag RR for total non-accidental hospitalizations peaked at lag 1 d after cold spell exposure, declined thereafter, and began to rise again after lag 5 d, reaching a second peak at lag 12–13 d before gradually decreasing. The lagged effects remained statistically significant during lag 8–18 d. The lag patterns of cold spell associations across disease categories, genders, and age groups were largely consistent with those of total hospitalizations.

Conclusion Cold spells have a significant impact on non-accidental hospitalizations of residents in Hengyang City, with notable lagged effects. The findings provide important theoretical support for establishing a more targeted cold spell early warning system.