Association of Seasonal Climate Variability and Age-Specific Mortality in Northern Sweden before the Onset of Industrialization

Abstract

Background and aims: Little is known about health impacts of climate in pre-industrial societies. We used historical data to investigate the association of temperature and precipitation with total and age-specific mortality in Skellefteå, northern Sweden, between 1749 and 1859. Methods: We retrieved digitized aggregated population data of the Skellefteå parish, and monthly temperature and precipitation measures. A generalized linear model was established for year to year variability in deaths by annual and seasonal average temperature and cumulative precipitation using a negative binomial function, accounting for long-term trends in population size. The final full model included temperature and precipitation of all four seasons simultaneously. Relative risks (RR) with 95% confidence intervals (CI) were calculated for total, sex- and age-specific mortality. Results: In the full model, only autumn precipitation proved statistically significant (RR 1.02; CI 1.00–1.03, per 1cm increase of autumn precipitation), while winter temperature (RR 0.98; CI 0.95–1.00, per 1 °C increase in temperature) and spring precipitation (RR 0.98; CI 0.97–1.00 per 1 cm increase in precipitation) approached significance. Similar effects were observed for men and women. The impact of climate variability on mortality was strongest in children aged 3–9, and partly also in older children. Infants, on the other hand, appeared to be less affected by unfavourable climate conditions. Conclusions: In this pre-industrial rural region in northern Sweden, higher levels of rain during the autumn increased the annual number of deaths. Harvest quality might be one critical factor in the causal pathway, affecting nutritional status and susceptibility to infectious diseases. Autumn rain probably also contributed to the spread of air-borne diseases in crowded living conditions. Children beyond infancy appeared most vulnerable to climate impacts.

On the association between weather variability and total and cause-specific mortality before and during industrialization in Sweden

“…As a proxy for Skellefteå weather, we used monthly data on precipitation and temperature from Umeå, a town 110 kilometres south of Skellefteå. For the period 1800-1859, early temperature recordings from Uppsala (roughly 600 kilometres south) and from Tornedalen (roughly 230 kilometres north) were used for the reconstruction of Umeå monthly temperature (see details in Rocklöv et al 2014). Weather data was obtained from the Swedish Meteorological and Hydrological Institute (SMHI), with the exception of Tornedalen data, which was taken from Klingbjer and Moberg (2003).…”

“…According to our earlier research, in Sweden between 1749 and 1859 annual average temperature and precipitation had little impact on annual death counts. But while higher springtime temperatures and precipitation decreased annual mortality, high levels of rain in autumn led to higher annual mortality (Rocklöv et al 2014;Schumann et al 2013). There was also evidence that older children were most vulnerable to an unfavourable climate, while there was a less pronounced impact on infants .…”

Importance of the Geocoding Level for Historical Demographic Analyses: A Case Study of Rural Parishes in Sweden, 1850–1914

“…For the proximity to wetlands variable, we estimate separate models for three high-risk groups: infants, children aged 1-5, and females aged 20-50 (WHO 2015). For the population density, we select children aged 1-15 who are sensitive to environmental factors (cf., e.g., Bengtsson 1999;Rocklov et al 2014). Separate models are estimated for the geocoding levels as well as for the different methods applied to the geographic context variables.…”

Unequal lands: Soil type, nutrition, and child mortality in southern Sweden, 1850-1914

“…Consequently, the differences in productivity changed from being regional to local in line with the enclosures. Finally, the effect of climate on mortality in preindustrial societies has been studied by combining microlevel and macrolevel data (Bengtsson and Broström 2010;Rocklov et al 2014;Schumann et al 2013). For example, Schumann et al (2013), using parish-level data, found that for a rural region in northern Sweden in the 18 th and 19 th century, a high amount of autumn rain increased the total number of deaths for all age groups.…”

“…Children aged 1-15 are studied because they were sensitive to malnutrition as well as to environmental factors (Bengtsson 1999;Rocklov et al 2014;Schumann et al 2013;Wolleswinkel-van den Bosch et al 2000) and the regional differences in mortality were usually larger among them (Van Poppel, Jonker, and Mandemakers 2005). As a sensitivity test, we also analyse infants (age 0-1) to compare the two age groups.…”