Heat adaptation measures in private households: an application and adaptation of the protective action decision model

The study addresses rising frequency and intensity of extreme heatwaves due to climate change and growing urbanization, which together increase heat-related health risks. The authors note substantial heat-related mortality globally and in Germany (e.g., over 20,000 excess deaths in 2018 in Germany by model calculation), and highlight calls for expanded adaptation measures, especially for nighttime heat exposure. Prior research indicates factors associated with heat adaptation behavior include risk perception, gender, age, income, education, and social networks. Despite this, adaptation behavior in German households remains relatively low and heat is often not perceived as a health risk, particularly among older adults who also report lower subjective heat stress. The research question is which factors influence the implementation of heat protection measures in private households. The study applies the Protective Action Decision Model (PADM) to heat adaptation behavior, leveraging measured indoor bedroom temperatures as an external stimulus (exposure) and testing hypotheses on how efficacy-related and resource-related attributes, heat risk perception, indoor temperature, and demographics affect adaptation behavior. The purpose is to identify determinants of household heat adaptation to guide communication and policy, and to propose adjustments to PADM based on empirical results.

The paper reviews theoretical frameworks used to study climate change and heat adaptation behavior. The Theory of Planned Behavior (TPB) posits behavior is driven by attitudes, subjective norms, and perceived behavioral control; it has been applied to climate adaptation and pro-environmental behavior, sometimes extended by moral obligation and socioeconomic/communication variables. The Value-Belief-Norm (VBN) theory is widely used for pro-environmental behavior but less for adaptation and lacks explicit risk perception. The Health Belief Model (HBM) explains preventive health actions via perceived susceptibility and severity (threat) and modifying factors; it has been applied to heatwave adaptation. Protection Motivation Theory (PMT) combines threat appraisal and coping appraisal (self-efficacy, response efficacy, cost) to explain protective intentions; however, inclusion of knowledge as a predictor is often needed. The authors argue that TPB, VBN, and HBM do not explicitly incorporate external stimuli (e.g., actual heat exposure), which is critical for heat adaptation studies. The Protective Action Decision Model (PADM) includes exposure and distinguishes efficacy-related attributes (e.g., perceived control) and resource-related attributes (money, time, knowledge). Given PADM’s prior success in other hazards, the study applies and adapts PADM to heat adaptation, formulating hypotheses about relationships among indoor temperature (as exposure), risk perception, efficacy/resource attributes, demographics, and heat adaptation behavior.

Study setting and design: The research was conducted in Augsburg, Germany (~300,000 inhabitants; July mean 18.1°C in 1981–2010; recent summers with more hot days). During July–September 2019, indoor temperatures were measured in private household bedrooms using Elitech RC-5 data loggers (±0.5°C accuracy) placed by participants who registered for the study. An accompanying German-language questionnaire collected data on heat risk perception, subjective heat stress (SHS), knowledge about heat risks, adaptation intentions and behaviors, demographics (age, gender, education, income, living alone), and health implications during heatwaves. A total of 431 datasets had both indoor temperature and completed questionnaires. Sampling and procedure: The city was partitioned by local climate zones; households along a cross-section were invited. Data loggers recorded every 15 minutes in July–August 2019. Survey participation was online or via telephone. Constructs were sourced from validated German instruments where possible; responses were translated to English for publication. Measures and constructs: - Heat adaptation behavior (HAB): Participants rated likelihood of five potential household measures (buy AC; install more shading; add green/blue areas; buy a fan; move to a cooler dwelling/city) and reported behaviors already used during heatwaves to ensure sleep quality (e.g., thin bedding/nightwear, nighttime showering, wet cloths, windows open at night, pre-sleep airing, sleeping in a cooler place). Items were dichotomized (option/already taken=1; not an option/not taken=0) and summed to an Adaptation Behavior Score (mean=6.15, sd=1.65; range 0–12). Descriptive uptake showed, for example, 78.9% considered/installing more shading; 62.9% buying a fan; 85.4% using thin/no bedding; 81.9% keeping windows open at night. - Heat risk perception (HRP): Based on validated items (Martens et al., GESIS), including statements about personal health risk and threats to plants/animals, rated on 4-point Likert scales, summarized into a score (Cronbach’s alpha=0.66). - Efficacy-related attribute (ERA): Internal locus of control (e.g., “I am in control of my own life”; “If I make an effort, I will succeed”), 5-point Likert scale, two items summarized (alpha=0.6). - Resource-related attributes (RRA): Financial resources (household income categories), time (employment status), and knowledge about heatwaves. Knowledge was measured via 10 true/false/I don’t know statements about heat risks (e.g., vulnerability of older/young people, heat-related diseases, bacteria growth, urban heat, nighttime stress). Correct answers were summed (alpha/KR-20≈0.45), categorized into low (0–4), moderate (5–7), and high (8–10). - Indoor temperature (INT): Mean indoor bedroom temperature during a 3-day heatwave (24–26 July 2019). Overall mean 27.14°C (sd=1.6). Distribution: 1.9% <24°C; 6.0% 24–25°C; 15.5% 25–26°C; 23.9% 26–27°C; 24.6% 27–28°C; 27.4% >28°C. - Subjective heat stress (SHS): Perceived stress at home during day and night (5-point Likert), summarized (alpha=0.71). Day: mean 2.71 (sd=1.11); night: mean 3.20 (sd=1.16). - Health Implication Score (HIS): Frequency of seven heat-related symptoms (1=never to 3=often), including drowsiness (M=2.10), sleeping problems (2.10), concentration problems (2.00), vertigo (1.56), headache (1.69), nausea (1.19), cardiovascular problems (1.44), summed. - Demographics: Age (<65 vs ≥65), gender, living alone; education and home ownership also described. Hypotheses (PADM-based): H1a: ERA positively relates to HAB. H1b: RRAs (income, time, knowledge) positively relate to HAB. H2a: Indoor temperature relates to ERA and knowledge. H2b: Indoor temperature relates to HRP. H3: HRP predicts HAB. H4: Demographics (age, gender, living alone) relate to HRP. Analytical approach: Descriptive statistics; Pearson correlations; hypothesis testing via one-way ANOVAs, linear regressions, t-tests; and a multiple hierarchical regression model (Step 1: variables initially directly linked with HAB in PADM; Step 2: all PADM variables; Step 3: additional variables SHS and HIS). Assumptions of normality and homogeneity were checked for ANOVAs (Shapiro-Wilk p>0.05; Levene p>0.05 where noted).

Descriptive and correlations: - Sample: N=431; 60.6% female; 15.5% aged ≥65; 58.9% held a university degree; 35.7% lived alone; income most commonly 1000–2000€ (21.1%) or 2000–3000€ (20.6%). - HAB correlated with: age (negative), SHS (r=0.337***), HIS (r=0.29***), knowledge (r=0.127**), HRP (r=0.18***), and indoor temperature (r=0.176***). Hypotheses tests: - H1a (ERA → HAB): Not supported by one-way ANOVA (p>0.05). However, ERA became significant in multivariate models (see hierarchical regression). - H1b (RRA → HAB): Income and employment not significant. Knowledge was significant: ANOVA F(2,462)≈4.88, p=0.008, f=0.21; Tukey post-hoc showed higher HAB for high knowledge vs moderate (Δ=−0.48; 95% CI [−0.88, −0.07], p<0.001) and vs low (Δ=−0.69; 95% CI [−1.36, −0.03], p<0.001). Supported for knowledge only. - H2a (INT → ERA, knowledge): Not supported; regressions not significant. - H2b (INT → HRP): Not supported (p=0.147). - H3 (HRP → HAB): Supported. ANOVA F(6,458)≈3.084, p<0.001, f=0.2. HAB increased from low to high HRP (Δ≈0.98; 95% CI [0.13, 1.84]) and from 2 to 4 HRP (Δ≈0.84; 95% CI [0.018, 1.66]). - H4 (Demographics → HRP): Supported for age only. Younger group had higher HRP than older (t(93.417)=2.839, p=0.006, f=0.14). Gender and living alone not significant. Hierarchical regression (Table 11): - Step 1 (R²=0.194; Adj. R²=0.038): HRP significant (β=0.142; p=0.013); ERA and RRAs not significant. - Step 2 (R²=0.405; Adj. R²=0.164): ERA significant (β=0.102; p=0.05); age (DCV1) negative (β=−0.284; p<0.001); indoor temperature positive (β=0.187; p=0.001). Direct effects of age and indoor temperature on HAB emerged. - Step 3 (R²=0.466; Adj. R²=0.217): Significant predictors: ERA (β=0.116; p=0.023), age (β=−0.209; p=0.001), indoor temperature (β=0.156; p=0.004), SHS (β=0.157; p=0.014), HIS (β=0.164; p=0.012). HRP not significant in this full model; knowledge not significant in the multivariate model despite earlier ANOVA support. Model adaptation: The adjusted PADM includes direct effects on HAB from self-efficacy (ERA), knowledge (supported in hypothesis testing), HRP (supported in hypothesis testing), indoor temperature, age, SHS, and health implications. Indoor temperature did not significantly predict HRP but directly predicted HAB.

Findings suggest that while univariate tests did not support a direct association of self-efficacy (ERA), income, or time with adaptation behavior, knowledge and risk perception are associated with greater adaptation. In multivariate models, self-efficacy, age, indoor temperature, subjective heat stress, and health implications significantly predicted adaptation behavior. Notably, indoor temperature did not significantly influence heat risk perception as hypothesized by PADM, indicating that measured exposure may not translate directly into perceived risk, at least in this context. Demographics influenced risk perception only through age; gender and living alone were not significant, differing from some international studies, possibly due to cultural or contextual differences. The direct effects of indoor temperature, SHS, and health symptoms on adaptation indicate that households tend to autonomously adopt measures when they experience heat burden or health impacts. For practice, targeting communication and interventions to older adults and enhancing public knowledge can foster adaptation. For policy, incorporating indicators such as indoor temperature, SHS, and health symptoms into municipal heat action plans can help set action thresholds and prioritize support, including for care institutions and community networks. The adjusted PADM proposed by the authors reflects empirical pathways identified in the hierarchical regression, providing a refined framework for future heat adaptation research and interventions.

The study applies and adapts the Protective Action Decision Model to explain heat adaptation behavior in private households, combining observed indoor temperatures with psychosocial and demographic factors. Key contributors to adaptation are self-efficacy, age (younger more likely to adapt), indoor temperature (higher temperatures associated with more adaptation), subjective heat stress, and experienced health implications; knowledge and heat risk perception are also relevant based on hypothesis testing. Importantly, adaptation behavior does not depend on higher income or time availability in this sample, suggesting broad-based potential for uptake if awareness and efficacy are addressed. Practically, authorities should focus communication and support on older populations and enhance knowledge about heat risks and effective measures. Methodologically, the PADM proved suitable but benefited from including direct exposure (indoor temperature), SHS, and health symptoms; future research should further validate the adjusted model, refine the knowledge construct for reliability, and replicate in other cities and cultural contexts to improve generalizability.

The sample may not be representative of all German cities; participants self-selected by volunteering to host data loggers and may be more sensitized to heat risks or have specific bedroom temperatures, education, or health profiles. Online survey timing across July means participants may have experienced different ambient conditions during response. The survey was conducted in German with some constructs adapted from English without a prior pilot, raising potential translation or measurement issues. Reliability of several constructs was below commonly desired thresholds (Cronbach’s alpha: HRP 0.66; ERA 0.6; knowledge 0.45; SHS 0.71), particularly for the knowledge scale measuring diverse facets; no pre-test was performed. Household-level responses likely reflect the views and behaviors of the respondent rather than all household members. Data are not publicly available yet due to ongoing publications. These factors may affect interpretation and generalizability.