Design, study population, and information sources

A cross-sectional study of the vulnerable population affected by energy poverty in Barcelona city participant of EJUSTA program or the non-energy poor population of Barcelona (from now on BCN) to compare the health status.

Two sources of information were used: 1) the EJUSTA Survey database, contain the baseline information of the EJUSTA participants, those were recruited consecutively at public social service offices during the 5-months (March to July 2016) of the execution of the program, and 2) the database of the Barcelona Public Health Survey (BPHS) 2016 edition. BPHS provides information from a representative sample of non-institutionalised Barcelona population obtained with a multi-stage random sampling. The questionnaire used was adapted from the BPHS using the same validated items.²⁶ So, identical health measures were included in both surveys along with demographic, socioeconomic, housing and energy poverty questions. Trained independent interviewers delivered the survey in both population face to face (only one member per household).

For the present study, EJUSTA participants that wished to receive free home energy counselling visits were also invited to participate in the research and their informed consent obtained. Individuals who declined to participate in the study also received the intervention. Inclusion criteria were ≥18 years, signed the informed consent and affected by energy poverty based on the consensual approach:²¹ inability to keep the home warm or inability to pay utility bills on time or presence of dampness and rot in the dwelling. Of 3094 participants, 2470 met the selection criteria (204 did not sign the informed consent; 248 were non-energy poor; 172 unknown energy-poor condition). The BCN individuals were sampled from the BPHS following the criteria of ≥18 years and people not affected by energy poverty using the same energy poverty measures but in the reverse sense. Therefore, individuals with one or more energy poverty indicator was excluded. Of the 4000, 2608 met the criteria (560 were <17 years, 799 reported at least one energy poverty measure, 33 unknown energy-poor condition).

The Clinical Research Ethics Committee has approved the study protocol of Parc de Salut Mar (Number 2016/6681/I).

Study variables

Four physical and mental health outcomes were analysed in this study. Self-reported health was measured using the question “How is your health in general?”, and the answer was categorised as good (excellent, very good, and good) and poor (fair, and poor). This has proven to be a good indicator of health status, health service utilisation and, mortality.²⁷ Presence or absence of asthma, chronic bronchitis and depression and/or anxiety in the previous 12 months were measured. Among the self-reported morbidity indicators, respiratory morbidity and the mental health disorders stand out for the high concordance between the registers and the self-reported values.^28,29

The main independent variable was energy poverty measured by three self-reported questions:²¹ “Could your household afford to keep the dwelling heated at an appropriate temperature during the winter months?”, “Has your household been in arrears on utility and community bills in the past 12 months?” and “Does your dwelling have any leaks, dampness on the walls, floors, ceilings or in the foundations, or rotten floors, windows or door frames?”. The answers were combined to create two variables: 1) “EP (de)composition”: a variable with seven categories based on all the combinations of the EP measures. 2) “EP intensity”: a 3-point categorical variable: “low EP” (yes in one measure); “moderate EP” (yes in two of them); and, “high EP” (yes in all).

Relevant covariables included were age, sex, place of birth (Spain and EU-15; outside the EU-15); household composition (one-person household; couple without children; couple with children; single parent; other); educational level (less than primary; primary; secondary or more); employment status (employed; unemployed; retirees and pensioners; permanent disability); and ability to face an unexpected financial expense of €750 or more (yes/no) that provide information about of individuals' preparation against financial risks.³⁰ Tenure status (owner; tenant; social housing) was also measured because of its relationship with dwelling conditions.

Statistical analysis

A descriptive analysis of all the variables in EJUSTA and BCN samples was conducted. Pearson's χ² test tested the significance of the differences between the groups. The prevalence of age-standardised health outcomes was calculated for both groups using the direct method, with the age distribution of the EJUSTA sample as a reference in both populations. Next, the prevalence ratios (PR) and their 95% confidence intervals (95%CI) for all health outcomes were estimated using Poisson regression models with robust variance adjusted by age.

As a sensitivity analysis, all the analyses were repeated, selecting non-energy poor manual class adult population from BPHS. Social class was measured using the Spanish Society of Epidemiology's classification,³¹ which is based on the current or last occupation. Then, the responses were classified as a non-manual class and manual class. We sampled the manual class population without any energy poverty measures. Results can be found in Table I in online ^Appendix .

Among the EJUSTA sample, a descriptive analysis of the energy poverty measures was performed. Then, the prevalence of age-standardised health outcomes by energy poverty intensity level we calculated. The association between poor health outcomes and energy poverty intensity was analysed by estimating the PR and their 95%CI for all health outcomes by using Poisson regression models with robust variance. The included explanatory variables were all covariables, except the inability to face an unexpected expense of €750 or more, because of collinearity with energy poverty intensity. Orthogonal polynomial contrasts were used to test the gradient association between poor health outcomes and energy poverty intensity.

All the analysis was carried out stratifying by gender, using the statistical software STATA13.

Health status among EJUSTA and non-energy poor population

The vulnerable populations studied reported severe adverse physical, and mental health outcomes compared to the non-energy poor population. The results support the findings of previous studies showing that low-income people, particularly women,²⁴ lower social classes, immigrants, and tenants, are more exposed to the structural determinants of energy poverty and its health consequences.^14,20 As previously noted, the household's inability to ensure the required levels of domestic energy services tends to coexist with other problems,^11,12 like financial difficulties, unemployment, and residential insecurity. The population studied was a socially disadvantaged subpopulation whose health is likely to worsen as more and more social problems coexist. There is abundant evidence of the association between health and socioeconomic status.³² However, this does not fully explain the association found between energy poverty and poor physical and mental health. The sensitivity analysis performed, as well as the energy poverty assessment by intensity level, confirmed this association. Both results support the conception of energy poverty as a condition related, but different from other socioeconomic vulnerabilities.³³ These results were consistent with the evidence found in previous studies related to asthma,^15,17 chronic bronchitis,¹⁶ and poor self-perceived health and mental health problems,^14,19 suggesting the problem goes beyond the mere existence of economic poverty or material deprivation.

Association between energy poverty intensity and health status among EJUSTA

Among the EJUSTA population, the self-reported energy poverty indicators showed a pervasive but uneven overlap in their dimensions. It was noted that the higher the energy poverty intensity, the higher the prevalence in almost all health problems studied. However, the energy poverty affects health through different pathways. For example, respiratory problems have been related to poor dwelling conditions.¹⁷In contrast, mental health problems have been more closely linked to worry about debt and affordability, thermal discomfort and concern about its health consequences.¹⁹ Such a suggestive pattern merits further exploration, not only about energy poverty intensity patterns suggested but also how the different pathways combined affect people's health. In future research, this approach should also address the inability to keep the home cool in summer, a problem that may become relevant in coming years given the climate change but remains relatively under-explored within EU.³⁴

Limitations and strengths

The main limitation was the existence of multiple confounding factors between the target and comparison populations, threatening its external validity. Although extrapolation of the health burden related to energy poverty should be done with caution, the evidence gathered from the analysis of the energy poverty intensity suggests an unequivocal and consistent relationship with several health outcomes. The sample was useful and a good first approach, but the comparison group far from ideal. Another limitation is the reliability and accuracy of the self-perceived energy poverty measures used due to possible reporting bias. For example, households that would be empirically identified as energy-poor with other metrics, may not identify themselves as energy-poor or may deny their circumstances.³³ In any case, the measures were asked identically in both populations. Lastly, this study may show a selection bias due to the non-random sample used among EJUSTA participants. For example, the sample did not include people in extreme socioeconomic circumstances affected by energy poverty who were not social service users (e.g. illegal immigrants, squatters, or people in extreme poverty). If these population groups had been included, the health outcomes would probably have been worse and the health associations even more substantial.

Despite these limitations, this is the first study analysing the association between energy poverty intensity and health in the context of Southern Europe. The advantages of a composite indicator that compares energy poverty level across households its ability to capture a multidimensional concept that often is appraised on separate components. and poorly studied in relation with health status. This composite indicator, although “rudimentary”, has allowed revealing how the energy poverty level was associated in gradient-shaped with different health outcomes. This approach may offer to policymakers the opportunity to design programmes according to the level of energy poverty, considering its possible associated health consequences.