The Warm Homes for Elder New Zealanders (WHEZ) study is a community-based trial designed to test whether an electricity voucher paid into the electricity account of older people with COPD, with the expressed aim of enabling them to keep their homes warm, results in reduced exacerbations.
Ethical consent was obtained from the New Zealand Multi-Region Ethics Committee MEC/07/05/062 and the Clinical Trial Registration is NCT01627418. Written informed consent was gained from the participants.
The primary research question is whether the provision of an electricity subsidy reduces exacerbations of COPD requiring hospitalisation, or treatment with corticosteroids and/or antibiotics. Secondary questions include whether the dwelling temperature is affected by the electricity intervention and whether the estimated benefits of the intervention exceed the costs.
The intervention is a subsidy of NZ $500 (approximately US $400, £250, €300) towards the household electricity bill, which was calculated as being sufficient for a 2.4 kW electric heater to run continuously for an additional 10 hours a day for 12 weeks, thus raising the temperature of one room substantially, as well as the overall average indoor temperature. The money was paid directly into the household’s electricity account (in the first winter of enrolment for the early group, and the second for the late group). $500 was considered a potentially feasible sum for the government to implement as a national policy and is comparable to the British Winter Fuel Subsidy.
The amount was tied to the electricity bill, as subsidising other common forms of heating, such as wood or liquefied petroleum gas, would potentially have a deleterious effect on indoor air quality, which might adversely affect the participants’ lung function. Initial feedback on this method of subsidy has been positive; there have been very few calls to our phone helpline from people expecting the payment to be made another way.
Due to the complexity of New Zealand’s deregulated domestic electricity industry, which includes competing electricity retailers, organising and implementing the intervention was not a trivial task. However, close liaison with the electricity companies proved worthwhile and most were very co-operative. For example, company representatives, who understood the study, referred back to the study team a request from a participant to withdraw some of the credited money from their account.
The primary outcome of concern was the rate of moderate or severe COPD exacerbations amongst the participants. The TORCH study of combined long-acting beta agonists and corticosteroids found yearly moderate or severe exacerbation rates dropped from 1.13 in the placebo group to 0.85 in the combination therapy group, giving a rate ratio of 0.75 [61
]. The study was sized to duplicate a reduction of this size – using a negative binomial model for exacerbation rates as recent analyses have used these, rather than Poisson models, as the negative binomial models better account for individual patient variability [61
]. A “k” parameter (the shape parameter of the associated gamma distribution) of 0.46 was used as a reanalysis of the TRISTAN data found this the most appropriate (95%CI 0.34-0.60) [62
]. These parameters, with a confidence level of 95% and a power of 80% yield an estimated required group size per arm of 239, which allowing for a 15% drop out rate (greater than the 10% found in the Housing, Insulation and Health study
for the over 65s due to the greater illness of this study population) suggested an initial sample size of 550 participants. If the “k” parameter was 0.4 then the required group size per arm would have been 230, if 0.6 then it would have been 262. If the baseline exacerbation rate was 1.0 rather than 1.13 then the required group size per arm would be 260. Therefore 550 participants was the target sample size.
From 2009 until December 2010, application forms were distributed through various agencies, including primary health organisations, COPD support groups, public hospital outpatient departments, and community health workers. Letters of invitation were also sent to eligible people discharged after a hospitalisation for COPD. Concerted attempts were made to follow up incomplete application forms, so that literacy difficulties did not automatically disqualify applicants. On receipt of an eligible application form, potential participants were sent an information sheet, personal consent forms for the person with COPD and another adult in the household (optional) and a reply-paid envelope. If the property was rented, consent forms were included for the landlord to agree to retrofitter assessment and insulation.
When the signed consent form was returned, the applicant was sent a letter confirming their acceptance into the study. Where necessary and feasible, ceiling and under-floor insulation were upgraded or retrofitted in homes.
Wave design and randomisation
As winter was the crucial period for the study, and recruitment is often slow at the start of the process the study was deliberately designed to recruit over two years in ‘waves’. Participants were recruited from three communities from 2009 to early 2011. Those recruited (i.e. consent and baseline data received) by June (early winter in the southern hemisphere) 2010 were assigned to Wave 1, those recruited later were assigned to Wave 2. Each wave was randomised into early and late intervention groups – the early group receiving the intervention the first winter in which the wave was enrolled in the study, and the late group the following winter. Thus Wave 1- Early participants received the fuel subsidy during 2010; Wave 1- Late and Wave 2- Early participants received the subsidy during 2011; Wave 2- Late participants received the subsidy during the winter of 2012. The randomisation was carried out by a statistician external to the study team.
The configuration of an autumn baseline with a spring follow-up was chosen in recognition of the significant illness of the sickest participants; with a full winter baseline there could have been more of a bias towards only the least unwell people completing the study. We allowed for a greater drop-out rate than we usually experience due to the illness among the study population.
As the intervention was carried out at a dwelling level, in order to avoid clustering, we selected only one primary participant per household, the person best able to complete the study questionnaire. The baseline interviews were carried out before randomisation so that knowledge of their group allocation could not affect the participants’ measures. In order to utilise the voucher for additional heating, the participants were then informed of their group allocation. Although the interviewers also knew the participant’s group allocation, they were trained to avoid mentioning group allocation status during interactions with the participants to minimise any recall bias. The analyses will be carried out with the researchers blinded to the meaning of the intervention group code. We also included several indicators designed to measure each participant’s propensity to report positive change and will control for the placebo effect by conducting sensitivity analyses using these propensity to positivity scores.
The flow of participants through the study is summarised in Figure , and the wave design portrayed in Figure .
Flow of participants through the study.
Study timeline. The wave design and randomisation meant that the interventions were spread over three years.
The demographic characteristics of the 522 COPD participants enrolled are presented in Table ; 20% were Māori, 77% New Zealand European or European and 2% Pacific peoples, and 3% reported belonging to other ethnic groupings, with only 3% of people reporting more than one ethnicity. We used the ethnicity classifications recommended by Statistics New Zealand. We succeeded in recruiting an older population (median age 71), with nearly a fifth (18%) of participants being over 80 at baseline.
Baseline characteristics of participants
Initially the intention was to accept people into the study whose post-bronchodilator spirometry results demonstrated GOLD stage II (moderate) COPD or worse (GOLD, 2005). However, the logistics of conducting high quality spirometry in the homes of COPD patients created a recruitment bottleneck, stalling the study progress. Instead, two extra questionsa
were added to the application form to identify people who had suffered a ‘moderate’ or ‘severe’ exacerbation of COPD as defined by the ATS/ERS Taskforce [63
], and were therefore considered at risk of subsequent exacerbations. Thereafter, those who had been admitted to hospital for COPD and/or taken steroids or antibiotics for COPD within the last three years were considered eligible.
Other essential requirements were for applicants to be living in the community, in a household responsible for its own electricity bill, and with the intention of remaining in the same house for the duration of the study. In recognition of the earlier onset of COPD morbidity and mortality for Māori, study participants had to be 55 years of age or older, although applicants of 45 years or over were considered if referred on the specific recommendation of community health workers.
Co-morbidities were not exclusion criteria, unless the co-morbidities rendered the participant unable to communicate effectively. Co-morbidities are to be expected in this age group and, to be effective, an intervention must improve COPD in the presence of co-morbidities. We used a Charlson style index [64
], modified to be based on self-report rather than ICD codes, to measure co-morbidities. Forty-two percent of baseline respondents reported having been told that they had two or more of: heart disease, cancer, liver disease, renal disease, diabetes and arthritis; 47% reported heart disease.
Using fuels other than electricity for heating was not an exclusion criterion, but participants were informed that only electricity bills would be credited. As we were attempting to model a possible population–level intervention, we wanted to have as broad inclusion criteria as feasible.
All participants were encouraged to be vaccinated annually against influenza, but this vaccination had to be obtained through their own GP or respiratory physician as part of the standard health-promotion campaigns. Eighty-four percent of baseline participants reported having had a flu injection for the most recent flu season and 22% reported a pneumococcal vaccination in the last five years.
In 171 households a support person also provided baseline information. These people completed a short questionnaire on the burden (if any) of providing care for the COPD participant, as well as the carer’s general health.
The study was set up in three predominantly low-income urban areas. This study was carried out in partnership with local organisations with an interest in health and housing [44
]. These organisations included a Māori health provider, a District Health Board, and Primary Health Organisations. Two communities were in the lower North Island (Whanganui and Wellington) and one in the South Island (Christchurch). Contracts were negotiated with these providers, who then worked closely with the research team, employing local field workers, nurses or community nurses, trained to conduct spirometry and administer the questionnaires.
One of the community organisations, Tu Kotahi, a Māori Asthma and Respiratory provider operating under the umbrella of Kokiri Marae, a large urban-based marae (communal meeting place), has participated in our two previous community housing trials. Our partnership is seen as resulting in many positive outcomes for both parties. For the Tu Kotahi Māori community it has brought insulated homes, improved heating, minimising the effects of allergens and increased information about healthy housing; for the University researchers it has provided access to the considerable knowledge of the community, as well as capacity building for both individual researchers and research groups.
Tu Kotahi’s involvement in housing research projects has highlighted the effects of poor housing on the health of their whānau (extended family community). In 2007, Tu Kotahi established a Māori COPD support programme and an important aspect has been the focus on healthy housing. Being research partners in this study has enabled Tu Kotahi to work more closely with whānau in their homes to assess their health status prior to, during and following the winter period, which has brought benefits beyond the study participation.
“The opportunity to follow up utilising our own community health worker has been crucial to maintaining the majority of whānau on the study. We have experienced a transient population during the study and this has often meant that we have only been able to locate whānau with information provided by the community health worker’s networks. Often, we discover that [members of the] whānau have been hospitalised or they have had to travel home for tangihanga (funerals). We also discovered how socially isolated a number of whānau are, this sometimes involved numerous visits before the door was opened for the community health worker to complete the study questionnaire and tests. We were also saddened at the number of whānau who passed away during the study; many of the whānau who participated experienced other co-morbidities that impacted on their respiratory conditions.”
The South Island provider, Community and Public Health, is the public health unit for the region of Canterbury. Staff also had a well-established working relationship with the researchers. Since 2001, these staff also carried out their own local health promotion strategies connected to housing improvement. Working with the researchers was therefore considered a mutually worthwhile undertaking in terms of developing opportunities for public health action.
In September 2010 the first of a continuing series of earthquakes and aftershocks occurred in Christchurch and the surrounding area [67
]. With considerable fortitude the Christchurch team did not let the earthquake, aftershocks or widespread liquefaction disrupt the spring visits by the field workers. In February 2011 a major earthquake caused a massive collapse of buildings and 185 lives were lost [68
Workplaces in the CBD were severely damaged and the area was closed to the public. The city’s infrastructure was in chaos as hundreds of organisations and businesses had no premises to work from, including the staff of Community and Public Health. This had an impact on the WHEZ study in a number of ways and the community researchers set about trying to overcome the subsequent problems.
Researchers agreed it was necessary to postpone interviewing for a three-week break to re-establish basic infrastructure and solve serious logistical problems. Laptops and other research equipment were initially out-of-bounds in the damaged office and only salvaged much later; however, the main computer system remained intact and data could be accessed remotely. The researcher’s home served as the research hub for the next year. There was only minimal disruption to the household visits when a further round of massive aftershocks and liquefaction occurred in June 2011. While the seismic activity has continued, it has diminished in intensity.
There was damage to the homes of the field workers and many participants, including chimney damage making fireplaces unusable, and structural damage affecting homes’ weather-tightness. There was also significant damage to infrastructure such as water, power lines and sewerage. For instance, because of damaged sewer lines in her neighbourhood, one field worker had to rely on a portable toilet on her street front for many months.
Some participants in the WHEZ study vacated their houses to a safer place for a period of time. Some moved to the less affected western suburbs or moved out of Christchurch for a number of weeks until they felt it was safe to return and this made it difficult for field workers to track participants down. Driving to the participants’ homes became laborious due to the extensive damage to roads. The study protocol stating that anyone who moved could not remain in the study was modified for the Christchurch participants; because so many other housing related variables had changed for many of the participants, there was little point in insisting on address stability. We also did not think it ethically appropriate to withdraw the fuel subsidy from participants who were forced to move because of the earthquake, but were otherwise keen to continue with the study.
The earthquake and continuous aftershocks were an emotionally traumatic experience for the participants, field workers and the co-ordinator. Financial worries about damaged housing, fear, and uncertainty about personal safety all contributed to this. The arrival of winter heightened people’s fears about the adequacy of the damaged electricity network to supply enough energy for households to be warm in winter. The field workers commented on the fears expressed to them during the research visits. The authorities were contacted for help in some desperate cases of damaged homes and heaters. Over the region, 9,000 new heaters were installed in the homes of people who had lost their primary heat source, and were paid for by the householder’s insurance company.
Spring and autumn questionnaires
The first main data collection commenced in autumn (2010 for Wave 1, 2011 for Wave 2). The participants’ basic anthropomorphic measures were collected, spirometry conducted and a health questionnaire completed. Height was measured using portable stadiometers (Seca 214), and weight with Tanita TIHD316 scales.
Pre- and post- bronchodilator spirometry was carried out, following ATS/ERS 2005 Task Force Guidelines [69
], using a portable ultrasonic spirometer (Easy One™). Subjects performed the tests seated. All community workers were trained to perform spirometry to ATS/ERS standards. FVC, FEV6
, PEFR, and FEF25-75
were recorded. Participants were encouraged to perform three acceptable manoeuvres, within a maximum of eight attempts. Participants who refused to attempt spirometry were not excluded from the study. Post-bronchodilator spirometry was conducted 30 minutes after the administration of 400 micrograms salbutamol (Ventolin®) via a Volumatic® spacer.
The questionnaire for the primary participant included sections on respiratory health, health care usage, general health, co-morbidities and quality of life, as well as tenure, heating used in the dwelling and house quality.
Similar data, except for height and weight, were collected during subsequent spring and autumn data collections. One or more data-loggers (HOBO or iButton®) monitoring the temperature (and, for a subset of dwellings, humidity) in the dwelling were placed in winter and collected in spring.
If another adult was living in the dwelling they were also invited to fill in a shorter ‘support person’ questionnaire, including sections on respiratory health, general health and the effects on caregivers of caring for the person with COPD on their life.
Summer and winter diaries
Participants were asked to fill in over two-week periods during summer and winter, a detailed diary asking about activity, symptoms, medication, and medical care used.
Objective data on health care usage were found using participants’ National Health Index (NHI) to access hospitalisations. Primary care data were obtained from Primary Healthcare Organisations directly and via the Healthlink network. Objective assessments of electricity usage were obtained from electricity companies (Table ).
Acceptability of methods The overall feedback from the majority of the participants in the WHEZ Study has, to date, been very positive. One such indicative report from a Wellington interviewer described her assessment of participants’ experiences in more detail:
“Although at the initial visit some participants had concerns and were unsure about the study, once they were given more information and further explanation of the study then they willingly and happily participated. In the first year pre-winter visits most of the participants reported how cold and damp their homes were and with the escalating cost of electricity they had to minimize how often they could have the heaters on to warm their homes. In the post–winter visit, the first wave of participants that received the subsidy of $500 commented how grateful and thankful they were for the subsidy because it helped tremendously - especially with the high winter electricity power bills. They also reported on the change in warmth in their homes since they had been insulated.”
Generalised linear models will be fitted to the data. Covariates potentially included in the model will be study arm, current medication, age, sex, region, baseline severity, smoking status and immunisation status (influenza and pneumococcal). The appropriateness of a Poisson or negative binomial model of exacerbations will be confirmed by examining the residuals of the observed data fitted to the postulated models against an envelope of residuals from a set of simulated trials. If the observed residuals fall inside the simulated envelope then the model is considered to fit adequately. A non-parametric test – rank analysis of covariance – will be used to confirm differences between the groups.
The design of a cost-benefit analysis of this study is straightforward and follows similar studies’ methods [70
]. The cost side of the intervention follows from providing a $500 voucher for electricity (and knowing the amount of electricity actually used). On the benefit side, the principal quantifiable outcome is reduced hospitalisations (ideally, exacerbation-reduced or exacerbation-free years would be measurable, but this was beyond the scope of the study). In addition, the study will estimate and value changes in pharmaceutical usage.