Developing a weighting strategy to include mobile phone numbers into an ongoing population health survey using an overlapping dual-frame design with limited benchmark information

Creating local estimates from a population health survey: practical application of small area estimation methods

Regular health surveys can produce reliable estimates at higher geographic levels but not for small areas. Alternatives are to aggregate data over several years or use model-based methods. We created and evaluated model-based estimates for four health-related outcomes by gender, for 153 Local Government Areas using data from the New South Wales Population Health Survey. The evaluation examined evidence on bias and determined the covariates available and appropriate for each outcome variable. The evaluation considered the likely precision of the resulting estimates. The bias and precision of results for single years (2006–2008) for each outcome variable using six covariate specifications were compared with direct survey estimates based on a single year's data and those obtained by aggregating over seven years. A practical issue is how to choose covariates to include in the models as the best covariate specification varies between outcome variables. Model-based results had median root mean squared errors between 3.3% and 5.5% (max 5.2% and 11.3% respectively) and median relative root mean squared errors between 6.8% and 24.5% (max 11.7% and 41.5% respectively). The model-based estimates were unbiased compared with direct estimates based on one or seven years of data and when aggregated to a point where direct estimates were reliable. The bias and reliability assessment process provides a way for policymakers to have confidence in model-based estimates.

Factors Associated with Home Fire Escape Plans in New South Wales: Multinomial Analysis of High-Risk Individuals and New South Wales Population

The preparation and practice of home-escape plans are important strategies for individuals and families seeking to reduce and/or prevent fire-related injury or death. The aim of this study was to assess the prevalence of and factors associated with, home-escape plans in the state of New South Wales (NSW), Australia. The study used data from two surveys—a 2016 fire safety attitudes and behaviour survey administered to high-risk individuals (n = 296) and a 2013 NSW health survey covering 13,027 adults aged 16 years and above. It applied multinomial logistic regression analyses to these data to identify factors associated with having a written home-fire escape plan, having an unwritten home-fire escape plan and not having any home-fire escape plan. The prevalence of written home-escape plans was only 4.3% (95% CI: 2.5, 7.5) for the high-risk individuals and 7.9% (95% confidence interval [CI]: 7.3, 8.6) for the entire NSW population. The prevalence of unwritten escape plans was 44.6% (95% CI: 38.8, 50.5) for the high-risk individuals and 26.2% (95% CI: 25.1, 27.2) for the NSW population. The prevalence of no-escape plan at all was 51.1% (95% CI: 45.2, 56.9) for the high-risk individuals and 65.9% (95% CI: 64.8, 67.1) for the NSW population. After adjusting for other covariates, the following factors were found to be significantly associated with unwritten-escape plan and no-escape plan prevalence: speaking only the English language at home, practicing home-fire escape plans infrequently, being married, being female and testing smoke alarms less often. Future fire interventions should target people who speak only English at home and people who test their smoke alarms infrequently. These interventions should be accompanied by research aimed at reversing the trend toward use of more flammable materials in homes.

Socio-demographic predictors of residential fire and unwillingness to call the fire service in New South Wales

In most industrialised countries, the majority of fire-related deaths and injuries occur in the home. Australia has implemented fire prevention programs and strategies, including the use of smoke alarms, to minimise this burden. The number of reported house fires has declined over the past decade. However, there is a growing recognition that unreported fires are important in the estimation of total fire hazards and their associated injuries. This current study used data from the 2014 New South Wales (NSW) Population Health Survey, a yearly telephone survey, consisting of 14,732 survey respondents. Univariate and multiple binary logistic regression models were conducted to examine predictors of residential fire and (un)willingness to call the fire service in the event of a residential fire. The proportion of respondents who experienced residential fires in NSW was 10% (95% confidence interval [CI]: 9.3, 10.8). The proportion of respondents who were willing to call the fire service was 3.1% (95% CI: 2.7%, 3.6%) and that of respondents unwilling to call was 6.9% (95% CI: 6.3%, 7.6%). Multivariate analyses revealed that respondents spoke another language in addition to English were significantly less likely to have experienced a home fire (odds ratio [OR] = 0.46; 95% CI: 0.32, 0.65, p < 0.001) and significantly less likely to call the fire service (OR = 0.34; 95% CI: 0.21, 0.54, p < 0.001), compared with those who only spoke English at home. The results in this study will inform Fire & Rescue NSW's ongoing development of appropriate interventions and awareness-raising programs about residential fire prevention.

•

Unreported fires are important in the estimation of fire hazards and their injuries.

•

In New South Wales 10% of population had reported that they had a residential fire.

•

For those who had residential fire, about 70% were unwilling to call fire brigade.

•

People who spoke other languages plus English were two thirds less likely to call.

•

Future interventions should target NSW residents that spoke other languages.

The impact of the mode of survey administration on estimates of daily smoking for mobile phone only users

Background

Over the past decade, there have been substantial changes in landline and mobile phone ownership, with a substantial increase in the proportion of mobile-only households. Estimates of daily smoking rates for the mobile phone only (MPO) population have been found to be substantially higher than the rest of the population and telephone surveys that use a dual sampling frame (landline and mobile phones) are now considered best practice. Smoking is seen as an undesirable behaviour; measuring such behaviours using an interviewer may lead to lower estimates when using telephone based surveys compared to self-administered approaches. This study aims to assess whether higher daily smoking estimates observed for the mobile phone only population can be explained by administrative features of surveys, after accounting for differences in the phone ownership population groups.

Methods

Data on New South Wales (NSW) residents aged 18 years or older from the NSW Population Health Survey (PHS), a telephone survey, and the National Drug Strategy Household Survey (NDSHS), a self-administered survey, were combined, with weights adjusted to match the 2013 population. Design-adjusted prevalence estimates and odds ratios were calculated using survey analysis procedures available in SAS 9.4.

Results

Both the PHS and NDSHS gave the same estimates for daily smoking (12%) and similar estimates for MPO users (20% and 18% respectively). Pooled data showed that daily smoking was 19% for MPO users, compared to 10% for dual phone owners, and 12% for landline phone only users. Prevalence estimates for MPO users across both surveys were consistently higher than other phone ownership groups. Differences in estimates for the MPO population compared to other phone ownership groups persisted even after adjustment for the mode of collection and demographic factors.

Conclusions

Daily smoking rates were consistently higher for the mobile phone only population and this was not driven by the mode of survey collection. This supports the assertion that the use of a dual sampling frame addresses coverage issues that would otherwise be present in telephone surveys that only made use of a landline sampling frame.

Electronic supplementary material

The online version of this article (doi:10.1186/s12874-017-0342-4) contains supplementary material, which is available to authorized users.

Profiling the mobile-only population in Australia: insights from the Australian National Health Survey

BACKGROUND

The Australian population that relies on mobile phones exclusively has increased from 5% in 2005 to 29% in 2014. Failing to include this mobile-only population leads to a potential bias in estimates from landline-based telephone surveys. This paper considers the impacts on selected health prevalence estimates with and without the mobile-only population.

METHODS

Using data from the Australian Health Survey - which, for the first time, included a question on telephone status - we examined demographic, geographic and health differences between the landline-accessible and mobile-only population. These groups were also compared to the full population, controlling for the sampling design and differential non-response patterns in the observed sample through weighting and benchmarking.

RESULTS

The landline-accessible population differs from the mobile-only population for selected health measures resulting in biased prevalence estimates for smoking, alcohol risk and private health insurance coverage in the full population. The differences remain even after adjusting for age and gender.

CONCLUSIONS

Using landline telephones only for conducting population health surveys will have an impact on prevalence rate estimates of health risk factors due to the differing profiles of the mobile-only population from the landline-accessible population.

Summary of the Impact of the Inclusion of Mobile Phone Numbers into the NSW Population Health Survey in 2012

Background

Although it was estimated that 20% of the population in Australia were mobile-only phone users in 2010, the inclusion of mobile numbers into computer-assisted telephone interviews (CATI) behavioural risk factor surveys did not occur until 2012.

Methods

Three papers have been published describing the methods, weighting strategy and the impact in detail of including mobile numbers into the NSW Population Health Survey (NSWPHS). This paper identifies the important components of those papers and summarises them for a broader audience.

Results

In the 2012 NSWPHS, 15,214 (15,149 with weights) interviews were completed (64% landline frame; 36% mobile frame). Response, cooperation and contact rates were 37%, 65% and 69% respectively. The inclusion of mobile phone numbers resulted in a sample that was closer to the NSW population profile and impacted on the time series of estimates for alcohol drinking, recommended fruit consumption, current smoking, and overweight or obesity.

Conclusions

The papers found that including mobile phone numbers into NSWPHS did not impact negatively on response rates or data collection, but it did cost more and affect the time series for some behavioural risk factors, in that it corrected the estimates that had been produced from a sample frame that was progressively getting less representative of the population.