Environmental burden of disease from unsafe and substandard housing, New Zealand, 2010–2017

Abstract Objective To assess the burden of disease related to unsafe and substandard housing conditions in New Zealand from 2010 to 2017. Methods We focused on substandard housing conditions most relevant for New Zealand homes: crowding, cold, damp or mould, and injury hazards linked to falls. We estimated the population attributable fraction using existing estimates of the population exposed and exposure–response relationships of health disorders associated with each housing condition. We used government hospitalization data, no-fault accident insurance claims and mortality data to estimate the annual disease burden from the most severe cases, as well as the resulting costs to the public sector in New Zealand dollars (NZ$). Using value of a statistical life measures, we estimated the indirect cost of deaths. Findings We estimated that illnesses attributable to household crowding accounted for 806 nights in hospital annually; cold homes for 1834 hospital nights; and dampness and mould for 36 649 hospital nights. Home injury hazards resulted in 115 555 annual accident claims. We estimated that direct public sector costs attributable to these housing conditions were approximately NZ$ 141 million (100 million United States dollars, US$) annually. We also estimated a total of 229 deaths annually attributable to adverse housing and the costs to society from these deaths at around NZ$ 1 billion (US$ 715 million). Conclusion Of the conditions assessed in this study, damp and mouldy housing accounted for a substantial proportion of the burden of disease in New Zealand. Improving people’s living conditions could substantially reduce total hospitalization costs and potentially improve quality of life.


Introduction
In New Zealand, dampness, mould and cold are common in both owner-occupied and rental dwellings. [1][2][3] In a government survey of 12 000 households in 2018-2019, 34% of New Zealanders reported that their homes were sometimes or always damp and 36% that their homes were mouldy. 4 Temperature measurements in approximately 6700 New Zealand homes in 2020 found that one third of homes had an average daytime inside temperature below 18°C. 5 In 2018, approximately half (388 310) of the 785 063 new injury claims due to falls in New Zealand happened in the home. 6 Much of the literature examining the relationship between housing conditions and health disorders worldwide has focused on specific adverse housing conditions. A review of the evidence linking housing conditions to health provided examples of local public health activities to address these issues. 7 However, we found few studies that assessed the total burden of disease from substandard housing conditions. [8][9][10] One study made a cost estimate of the total burden as part of a cost-benefit analysis of making the housing stock in England healthy and safe. 10 A study in the World Health Organization (WHO) European Region demonstrated that the environmental burden of disease approach is feasible for studying substandard housing conditions. 8 None of these studies, however, detailed the burden from each health disorder or analysed which disorders were the biggest drivers of costs.
Here we estimate the burden of disease attributable to four substandard housing conditions found most frequently in New Zealand: household crowding, cold housing, damp or mouldy housing, and injury hazards linked to falls. We provide policy-makers with information to understand where resources might best be targeted and the potential benefits of policies undertaken to improve poor housing conditions.

Methods
We used WHO methods of assessing the environmental burden of disease at the national level. 11 We selected the household risk factors to analyse (crowding, cold, damp or mould, and hazards leading to falls) based on new recommendations in the 2018 WHO Housing and health guidelines. 12 We did not consider housing conditions which had existing guidelines (toxic materials such as lead and asbestos; water, sanitation and hygiene; and indoor air pollution from solid fuels and noise) or where exposure in New Zealand is principally occupational (lead and asbestos). 13,14 Although New Zealand has a guideline for dampness and mould, we included this condition because lack of insulation and low indoor temperatures affect the extent of dampness and mould. 12 We focused on the aggregate burden of these household conditions because the solutions to these problems are not necessarily independent.

Data sources
The first step was determining the proportion of the population exposed to the studied household risk factors in New Zealand homes. For household crowding, we used data from the national census on the proportion of the population reported to live in crowded conditions. 15 For cold and dampness or mould, we obtained data from the New Zealand General Social Survey on the proportion of people reporting their home was colder than they would like and the proportion who had a problem with dampness or mould. 3 For exposure to risk of falls, we used data from a randomized controlled trial about the proportion Objective To assess the burden of disease related to unsafe and substandard housing conditions in New Zealand from 2010 to 2017. Methods We focused on substandard housing conditions most relevant for New Zealand homes: crowding, cold, damp or mould, and injury hazards linked to falls. We estimated the population attributable fraction using existing estimates of the population exposed and exposure-response relationships of health disorders associated with each housing condition. We used government hospitalization data, nofault accident insurance claims and mortality data to estimate the annual disease burden from the most severe cases, as well as the resulting costs to the public sector in New Zealand dollars (NZ$). Using value of a statistical life measures, we estimated the indirect cost of deaths. Findings We estimated that illnesses attributable to household crowding accounted for 806 nights in hospital annually; cold homes for 1834 hospital nights; and dampness and mould for 36 649 hospital nights. Home injury hazards resulted in 115 555 annual accident claims. We estimated that direct public sector costs attributable to these housing conditions were approximately NZ$ 141 million (100 million United States dollars, US$) annually. We also estimated a total of 229 deaths annually attributable to adverse housing and the costs to society from these deaths at around NZ$ 1 billion (US$ 715 million). Conclusion Of the conditions assessed in this study, damp and mouldy housing accounted for a substantial proportion of the burden of disease in New Zealand. Improving people's living conditions could substantially reduce total hospitalization costs and potentially improve quality of life.

Research
Burden of disease from housing, New Zealand Lynn Riggs et al.
of homes in need of repair to prevent falls among a sample of houses typical of New Zealand housing. 16 More details of the selection of risk factor exposures are presented in Table 1. The second step was obtaining data on health disorders associated with the studied housing conditions and the exposure-response measure (an odds ratio or relative risk) of each problem. We did not conduct a systematic review of the literature but used data from existing reviews or meta-analyses worldwide ( Table 2). For crowding, indoor cold and injury hazards we relied principally on the systematic reviews which informed the WHO housing and health guidelines. 12 For data on damp or mouldy housing, we used the WHO guidelines on dampness and mould. 27 We selected health outcomes where the certainty of evidence in the systematic reviews was rated as medium or high. We included only falls in our assessment of injury hazards from home injuries as the WHO housing and health guidelines grade the quality of evidence for hazards other than falls as low or very low. More details of our criteria for selection of health disorders and exposure-response measures for the analysis are provided in Box 1.
The third step was obtaining data on outcome measures. For the health disorders associated with household crowding, cold and dampness, we obtained individual-level data from the health ministry on publicly funded hospitalizations and deaths. 28 Specifically, we used administrative hospital admissions data from 2010 to 2017, including emergency department visits, for New Zealand residents where the primary diagnosis was associated with each of the studied health disorders. Diagnoses for each hospitalization are coded using the International statistical classification of diseases and related health outcomes, Tenth revision, Australian modification (ICD-10-AM). 29 We considered hospital admissions within 7 days of another hospital discharge as part of the same case. For these cases, we used the primary diagnosis from the first hospitalization to assign a specific health disorder, and if the diagnosis did not match our list of diagnosis codes we used the second hospitalization diagnosis, and finally the third hospitalization diagnosis. By this method we assigned all events to only one health disorder so that outcomes were mutually exclusive, and no events were left unclassified. We also obtained health ministry data on deaths from each studied health disorder from 2010 to 2014, as data were only available up to 2014. 28 These data included the underlying cause of death (coded using ICD-10-AM), age and ethnicity.
For home injury hazards, we obtained administrative claims data for medically treated injuries from the government's Accident Compensation Corporation which provides comprehensive coverage of the public-sector costs of accidents. Everyone in New Zealand is covered by this no-fault scheme if they are physically injured in an accident, with the scheme paying for medical treatment, lost wages, childcare, counselling, therapy and death benefits. We included claims for injuries between 2010 and 2017 where the scene of injury was listed as the home, the claimant  In Braubach et al., the IRR used for tuberculosis and crowding was 1.5 (plausible range of 1.2 to 2.0) for European countries with medium and high tuberculosis incidence. 8 We did not use this measure since it was not included in the WHO guidelines 12 and New Zealand is a low tuberculosis incidence country. This rate is provided for comparison. d To be conservative, we use the low humidity odds ratio to calculate the population attributable fraction; however, we provide all the odds ratios for comparison. e Respiratory infections category excluded any diagnostic codes in this chapter that are specified as part of the other health disorders. For example, J45 and J46 are included under asthma and hence are not included in the broader respiratory infections category. Therefore, the categories are mutually exclusive and we can aggregate the health outcome measures without double counting.
f Given that our outcome measures are hospitalizations, we expect that the current asthma odds ratio is the most relevant. The other odds ratios are provided for comparison.

(. . .continued)
Burden of disease from housing, New Zealand Lynn Riggs et al.
was a resident, and the classification implied a fall.

Data analysis
We estimated the population attributable fraction (PAF) for each health disorder using the following equation: where P is the proportion of the population exposed to the risk factor and RR is the exposure-response measure (odds ratio or relative risk) of the disease. This fraction represents the proportional reduction in adverse health outcomes that would occur if exposure to these risk factors was eliminated. Table 1 and Table 2 show the values of P and RR we used to calculate the population attributable fraction for indoor crowding, cold, dampness and mould. As children were over-represented in crowded households, we adjusted P for age when the associated health disorders were for specific ages. For falls, we used a population attributable fraction of 26% from other authors' estimates. 16 For health disorders linked to household crowding, cold or dampness or mould, we first determined the total length of stay and total cost of each hospitalization. Costs are in New Zealand dollars (NZ$), using 2017-2018 unit prices. The exchange rate during this time period was NZ$ 1 to 0.7148 United States $ (US$). We then aggregated these case-level estimates to estimate the number of hospitalizations, the number of unique patients, the total hospitalization cost and the total number of nights in hospital for each calendar year and then calculated the annual mean for each health disorder. For household crowding, some health disorders were specific to certain age groups. Hence, we used patient age to exclude hospitalizations outside the age range for these disorders. To estimate the burden of disease attributable to household crowding, cold and dampness or mould, we simply multiplied our population-level disease estimates by the population attributable fraction.
Using individual death records, we used the ICD-10-AM codes for the underlying cause of death to classify deaths into the same categories as hospitalizations. We then aggregated the individual records to tabulate the total number of deaths for each health disorder in each calendar year and then calculated the annual mean to obtain the populationlevel estimates. Next, we multiplied the population-level estimates by the population attributable fraction to estimate the deaths attributable to household crowding, cold and damp or mould.
For home injury hazards, we first used the claims data to estimate the population-level annual number of fatal and non-fatal fall injuries and the associated costs (such as medical treatment, lost wages, death benefits) from these injuries for each year in 2010-2017 by injury severity and then estimated the mean for this time period. The accident compensation scheme provided the data in six mutually exclusive injury severity categories for each claim. We used cases in the fatal injury category to estimate the number of deaths. Claim costs included all costs paid as at 1 October 2018 in NZ$. We then estimated the attributable burden by multiplying the population-level estimates by the population attributable fraction.
Finally, to estimate the societal costs of mortality attributable to all the studied household conditions, we used a willingness-to-pay based value of a statistical life. 30 We estimated the value of a statistical life using a Value of Safety survey conducted in 1991 which asked New Zealanders about their willingness to pay for safety improvements that are expected to avoid one premature death. 31 The 1991 value is indexed to average hourly earnings by the transport ministry to express it in 2017 NZ$. 31

Ethical considerations
Ethical approval for this research was granted by the University of Otago Ethics Committee (reference number HD18/094).
Access to anonymized data provided by Statistics New Zealand was in accordance with security and confidentiality provisions of the Statistics Act 1975. Only people authorized by the Statistics Act 1975 are allowed to see data about a particular person, household, business or organization and the results in this paper have been confidentialized to protect these groups from identification.
We carefully considered the privacy, security and confidentiality issues associated with using administrative and survey data in the Integrated Data Infrastructure. Further detail can be found in the Privacy impact assessment

Box 1. Selection of health disorders and exposure-response measures due to substandard housing in New Zealand
For determining which exposure-response measure to use when several studies were available, we selected studies based on three factors: (i) the quality of the research design; (ii) the exposure measure; and (iii) the similarity of the study population to New Zealand's population. When looking at the quality of the research design, we ranked studies in the following order (starting with highest quality): meta-analysis; randomized controlled trial; well designed controlled trial without randomization; well designed cohort or case-control study; well designed ecological studies. In some instances where studies were of similar quality, we selected an exposureresponse measure from a given study because the underlying population was more similar to New Zealand's. For dampness and mould, we included those health disorders with sufficient evidence of an association: asthma, upper respiratory tract infection, cough, wheeze, dyspnoea and respiratory infections. 27 In addition, we included bronchitis because a subsequent meta-analysis 25 indicated stronger evidence of an association between dampness and mould and bronchitis than was available for the World Health Organization (WHO) guidelines on dampness and mould. 26 Moreover, since our outcome measure was principally hospitalizations, we excluded disorders that cannot be clearly linked to hospitalizations. For example, the quality of the evidence on the relationship between cold indoor temperatures and blood pressure is rated as high. 12 However, since evidence on the relationship between cold and related hospitalizations was not available, we did not include blood pressure in the analysis. For crowding, selection of data was complicated by studies using different crowding measures and very different populations. For tuberculosis risk due to household crowding, we did not use exposure-response measures based on research design quality and instead used data from an ecological study since it was for New Zealand. Moreover, the cohort and case-control studies that we would otherwise select had higher effect sizes. Hence, we may have underestimated tuberculosis outcomes, but this would not greatly change our results. We also deviated from the WHO guidelines on housing and health 12 for upper respiratory tract infection and gastroenteritis due to difficulties in selecting appropriate studies. Instead, we used effect sizes from a metaanalysis for household crowding 17 that was not included in the WHO guidelines. 12 The effect sizes we used were generally lower than those found in the guidelines. 12

Population attributable fraction
The population attributable fraction for each health disorder linked to a housing condition is shown in Table 3. The highest population attributable fraction was for cough linked to dampness and mould (17.6%; uncertainty range: 12.7-22.5%) and the lowest was for tuberculosis from crowding (0.5%; uncertainty range: 0.2-0.8%).

Burden of hospitalization
The burden of disease estimates for household crowding, cold, and damp and mould are shown in Table 4. Overall, we estimated that 499 patients were hospitalized annually for illnesses attributable to household crowding (uncertainty range: 9-1722 patients). We estimated that 625 hospitalizations annually were attributable to living in cold homes (Table 4), accounting for 1834 nights in hospital and costing more than NZ$ 2.3 million (US$ 1.6 million; uncertainty range: NZ$ 1.1 million to 4.3 million). In this group, hospitalizations for cold or influenza had the highest cost at more than NZ$ 1.3 million (US$ 0.9 million). However, the same number of patients (260 patients) were hospitalized for wheeze as for colds or influenza but with 40% of the costs (NZ$ 0.5 million; US$ 0.4 million) and one quarter of hospital nights (243 versus 997 nights). Chronic obstructive pulmonary disease, on the other hand, accounted for 48 patients with 69 hospitalizations, but with costs similar to those for wheeze (NZ$ 0.5 million; US$ 0.4 million). This difference may be due to longer hospital stays for chronic obstructive pulmonary disease (8.6 nights per hospitalization) compared with cold or influenza (3.8 nights per hospitalization) or wheeze (0.8 nights per hospitalization).
We also found that more hospitalizations between 2010 and 2017 were attributable to dampness and mould than to cold or crowding (Table 4). In total, we estimated that annually 6276 hospitalizations of 5666 patients were attributable to dampness and mould accounting for 36 649 hospital nights costing almost NZ$ 36.0 million (US$ 25.7 million). Moreover, pneumonia or lower respiratory tract infection was the largest contributor to these totals, with 2486 hospitalizations annually for 28 253 nights costing approximately NZ$ 23.7 million (US$ 16.9 million). The second most costly condition was upper respiratory tract infections. Annual upper respiratory tract infection hospitalizations were about half of those for pneumonia or lower respiratory tract infection (1308 versus 2486 hospitalizations) but costs were about one tenth (NZ$ 3.2 million; US$ 2.3 million). This cost difference may be due to fewer nights in hospital; the average patient with pneumonia or lower respiratory tract infection spent 11.4 nights in hospital, whereas the average patient with upper respiratory tract infection spent just over one night. CI: confidence interval. a For some health disorders associated with household crowding, we use specific age groups. Hence, we adjusted the risk of exposure to account for the age group. See Table 1 for more details of the proportion of the population exposed. b See Table 2 for more details of the selection of exposure-response measures.

Burden of deaths
Overall, we estimated approximately 68 deaths were attributable to falls (Table 5), one death to household crowding, 16 deaths to cold and 145 deaths to dampness or mould (Table 6), for a total of 229 deaths annually. Using the value of a statistical life of NZ$ 4.2 million (June 2017 NZ$), 31 we estimated the total cost of these deaths due to unsafe and substandard housing conditions to be NZ$ 938.9 million (US$ 671.1 million) annually.

Discussion
We estimated that approximately 8300 hospitalizations (0.4% of nearly 2 million annual average hospitalizations over the same time period, 2010-2017) and 229 deaths each year were attributable to substandard or unsafe housing conditions in New Zealand. Dampness and mould imposed the largest disease burden in terms of hospitalizations and deaths compared with the other housing conditions we analysed. Comparing costs across all housing conditions is difficult, since we have more complete information for injuries due to the comprehensive nature of the claims data. Even so, claims costs for hospitalized injury cases (almost NZ$ 30 million)which include all costs paid in relation to the claim and not just the cost of hospitalization -were still exceeded by the estimated hospitalization costs attributable to damp and mouldy housing (around NZ$ 36 million).
Our total annual cost estimate includes only direct costs to the public sector through hospitalization costs and injury claim payments. We estimated the mortality costs separately but did not include other costs for morbidity such as productivity losses from missed work, other than the earnings paid through injury claims. However, even this value underestimates total lost wages since only 80% of earnings are covered by the accident compensation scheme. 32 From other studies, we also know that the total costs to society are likely to be much higher than the costs estimated here. For example, research valuing disability-adjusted life years using the value of a statistical life estimates the total cost of preventable home falls at around NZ$ 22 billion (in 2012 NZ$). 33 Comorbidities may exacerbate the impact of these housing conditions on health, and many homes have overlapping issues, meaning that the health disorders resulting from these housing conditions may not be simply additive. For example, many cold homes are also damp or mouldy. We reported out-  comes attributable to cold separately from damp and mould, even though the conditions are related, 34 because the exposure-response rates for the health disorders are linked to one condition. One report suggested that many New Zealand homes would experience far fewer periods of high relative humidity if they were heated to a minimum of 18 °C throughout. 34 However, we were unable to find any research that measured the interaction of these effects or the effect of comorbidities on these disorders. Hence, we did not account for the interaction of multiple housing conditions or comorbidities in our estimation of health disorders.
Other studies have attempted to assess the overall burden and societal costs from unsafe and substandard housing conditions. 9,10 An estimate of the costs of poor housing conditions in England showed that the most hazardous conditions in English homes cost the National Health Service (NHS) in excess of 600 million British pounds (£) per year, with the cost to society in excess of £1.5 billion per annum. 10 Updated cost estimates, published in 2015, showed a £1.4 billion per annum cost for the NHS, putting the cost on par with those from physical inactivity, smoking or alcohol. 9 Our estimates are of a similar magnitude given differences in the size of the two populations and currency exchange rates.
Our estimates of the disease burden are conservative since we do not have information about the number of nonhospitalized people for many of the housing-related health disorders we included (injuries are the exception), and hence, the costs do not include general practitioner visits or pharmaceutical costs. Moreover, the costs included in our analysis are primarily costs to the public sector from hospitalizations and generally do not include other social costs, except for mortality. For injuries, we included costs paid related to the claim (including wages paid for time off work), but these costs do not include other costs to society, only direct outlays.
Basing our estimates on self-reports of 21.2% of the population exposed to cold and 31.8% exposed to damp or mould may underestimate the true exposure as householders in New Zealand typically report better conditions than assessors' reports for the same homes. For example, in one survey of 560 houses, assessors reported visible mould in 49% of houses. 1 Other surveys recorded night-time temperatures below 18 °C in 84% of bedrooms (83 homes), 35 and daytime temperatures below 18 °C in 33% of homes (6700 homes). 5 Moreover, in homes with recorded daytime temperatures below 16 °C (15.1% of homes surveyed in winter), only 36% of householders thought their homes were always or often cold in winter and 45% were able to see their breath inside. 5 Nevertheless, subjective measures of poor housing have been consistently linked with a significantly increased risk of health effects. 36,37 We therefore believe that using self-reported or other subjective measures of dampness and mould for our exposure measure is supported by current research.
Our study is also likely to underestimate the true burden of poor housing since our data were limited to hospitalizations and deaths for most housing conditions (except fall injuries). For health disorders associated with the other housing conditions, most patients will never be hospitalized and some may never seek medical care, leading to undercounting of cases. For the direct costs to the public health-care sector, however, hospitalizations are likely to capture a substantial portion of the costs. In one study of housing-related health disorders for children, the majority of health care costs averted were due to reductions in hospitalizations, despite far greater numbers of general practitioner visits and prescriptions. 38 Our analysis has highlighted gaps in the evidence. Further research is needed to better understand the interaction effects of different housing conditions and the resultant health effects. Moreover, more work is required to fully estimate the total burden, including the total cost to society. Future work would also benefit from a focus on vulnerable populations (such as home renters, lowincome households) and the potential impact on inequality. Still, our estimates could be used to target policies at specific populations. For example, given the high cost of pneumonia or lower respiratory tract infection for household dampness and crowding, working to improve living conditions for these patients could substantially reduce total hospitalization costs and potentially improve their quality of life. новозеландских долларов (100 миллионов долларов США) в год. Было также определено, что с неблагоприятными жилищными условиями в общей сложности ежегодно связаны 229 смертей и ущерб для общества от этих смертей составляет примерно 1 миллиард новозеландских долларов (715 миллионов долларов США).