A cross-sectional study exploring equity of access to telehealth in culturally and linguistically diverse communities in a major health service

Study design and setting

This cross-sectional study used service activity data from the outpatient department of four tertiary hospitals in Queensland, Australia, from January 2021 to December 2021. This included metropolitan outpatient services from the Metro South Health Region (MSH): Princess Alexandra Hospital (PAH), Logan Hospital (LGH), Queen Elizabeth II Hospital (QEH) and Redland Hospital (RLH).

MSH serves a diverse community considered the most culturally diverse area in Queensland, with 20% of the population having a first language other than English.²² MSH also serves a population with low socioeconomic status, high levels of disability and a significant Aboriginal and Torres Strait Islander community.²² The diversity of the MSH population and the local technical support for service provision make this health region a suitable area to study access to telehealth services. We adopted the STROBE Statement Guideline for reporting observational studies.²³

Data collection and sample

This study examined outpatient department appointment data stored in the scheduling platform Enterprise Scheduling Management (ESM).

We used data from the top 10 services with the highest videoconference use across MSH to ensure the feasibility of the study, including Hepatology, Pulmonary Rehabilitation, Gastroenterology, Immunology (including Allergy), Psychology (including Neuropsychology), Speech Pathology, Cardiothoracic Surgery, Rehabilitation Medicine and Cardiac Rehabilitation. Registries with missing data related to the variables of interest (listed below), such as telehealth use or the use of English as a first language, were excluded from the analyses.

Variable measures

This study considered telehealth use as an independent variable, including appointments provided via telephone and videoconference. Appointment characteristics included the month of the appointment, the facility where the appointment was provided, clinical service (reported a corporate clinical code) and appointment type (e.g. telephone new, telephone review, videoconference new, videoconference review). Patient characteristics included age (years), sex (male/female/indeterminate), country of birth, Aboriginal and Torres Strait Islander status, interpreter requirement (yes/no) and whether English was the first language (English/non-English). Recognising the multifaceted nature of CALD consumers, we simplified its various dimensions into three different categorial variables: (1) country of birth, (2) use of English as a first language and (3) need of interpreter. This approach facilitated the comprehension of the intricacies associated with CALD consumers accessing telehealth. It allowed us to mitigate potential confounding effects related to other sociodemographic factors or modalities of care. Provided postcodes were categorised into deciles using the Index of Relative Socioeconomic Disadvantage (IRSD) as described by the Australian Bureau of Statistics.²⁴ A lower score indicates that an area is at a relative disadvantage compared to an area with a higher score.²⁴ In order to capture a level of cultural and linguistic diversity in one variable, we created a composite variable combining both country of birth and whether English was a first language. Therefore, we recoded ‘country of birth’ into three categories: ‘Australia’ for those born in Australia; ‘not Australia, not prioritised’ for people born in countries other than Australia whose first language is English; and ‘not Australia, prioritised’ for people born in countries other than Australia whose first language is not English. Aboriginal and Torres Strait Islander status was examined as a separate variable to enable specific description of Aboriginal and Torres Strait Islander peoples health service use (outside the scope of this study). Age was treated as a continuous variable. All other demographic variables were treated as either categorical or binary.

Statistical analyses

De-identified data from the electronic medical system were exported into a Microsoft Excel® CSV file and then imported into STATA^® SE 17 for the statistical analyses. Chi-squared tests were used to compare telehealth use/non-use and its modalities (phone versus videoconference) with patient sociodemographic factors reported to influence telehealth use, such as age, sex, use of English, country of birth, the requirement of an interpreter, Aboriginal or Torres Strait Islander status and IRSD. Univariate and then multivariate logistic regression was conducted using a panel set to group patient events together. Odds ratios (OR) for the association of telehealth use and sociodemographic factors were set, with a 95% confidence interval (CI) and a significance cut-off level <0.05. Similar models were set to assess the association between telehealth modalities and sociodemographic factors. The final multivariate adjusted regression models were adjusted for age, sex, use of English, the requirement of an interpreter and IRSD.

Ethics

The Metro South Research Ethics Committee approved the study (HREC/2021/QMS/81523).

Appointment characteristics

Our dataset included N = 153 427 appointments conducted in 10 clinical services during the observation period (Table 1). PAH represented 66.50% (n = 102 029) of the total number of observations included in this study (Table 1). Across the 10 clinical services included in the study, Gastroenterology contributed 45.19% of the appointments (n = 69 341).

Patient characteristics

Patients had a mean age of 53 years (s.d. ± 17.61), with a similar distribution of males and females (49.80%/50.17%). Regarding place of birth, 62.73% (n = 96 249) of patients were born in Australia, 15.19% (n = 23 299) were born outside Australia in an English-speaking country and 22.08% (n = 33 878) were born in countries whose first language was not English. Aboriginal and/or Torres Strait Islander peoples represented 4.10% (6218) of the total observations.

A total of 88.42% (n = 132 722) used English as their first language, while 11.58% (n = 17 393) used a language different from English. Only 6.09% (n = 9163) of the total sample were recorded as needing an interpreter for their appointments. The mean IRSD was 5.47 (s.d. = 2.89), indicating the population did not have a low socioeconomic background on average.

Summary of telehealth services

Telehealth accounted for 39.16% (n = 60 081) of outpatient consultations conducted across all observations. The most common telehealth modality was the telephone (64.87%, n = 38 514), followed by videoconference calls (35.13%, n = 20 857). Across the different facilities, PAH used telehealth for 42.54% (n = 43 402) of all consultations, followed by LGH with 28.27% (n = 8389), QEH with 42.42% (n = 6424) and RLH with 28.34% (n = 1866). Patients who used telehealth (52.35 years, s.d. = 17.97) were, on average, slightly younger than the ones not using telehealth (52.78 years, s.d. 17.05) (t-test = −4.63, P < 0.0001). There was no significant difference in telehealth use according to sex.

Telehealth use among CALD consumers

Among people not born in Australia and categorised as prioritised, telehealth use was 19.53% (n = 11 734) compared to 64.54% (n = 38 775) of telehealth use among patients born in Australia. Among the former, people who reported not using English and accessed telehealth represented 41.53% (n = 4806) of telehealth use among this group and 8.21% of the overall telehealth use. The frequency of telehealth use among the population born outside of Australia, not using English as their primary language and requiring interpreter services was 19.03% (n = 1675) of the total number of telehealth users (χ²(1) = 1.70, P = 0.0001). Among Aboriginal and Torres Strait Islander peoples, telehealth use was 37.86% (n = 2354) (χ²(1) = 4.02, P = 0.04). The level of socioeconomic disadvantage was slightly higher among telehealth users, with 5.72 (s.d. = 2.86), compared to the in-person appointments (5.31, s.d. = 2.89) (t-test = −26.94, P < 0.0001). While overall phone use was higher across all groups, stratified analyses per category showed different results. The proportion of use of videoconference among people born outside Australia in a prioritised country was significantly higher (41.10%, n = 4785, χ²(2) = 227.56, P < 0.001) compared to people born in Australia (33.79% n = 12 935) and those born in a non-prioritised country (33.19%, n = 3137). Significant differences between groups were detected among those needing interpreter services (40.70%, n = 689, P < 0.001) and those not using English as a first language (46.43%, n = 2330, P < 0.001). A detailed description of these findings is provided in Supplementary Table S1.

Regression analysis

After fitting multivariate panel logistic regression models and adjusting for confounders, our analysis showed no evidence for an association between telehealth use and age (OR 1.00, 95% CI 1.00–1.01, P = 0.0001) (Table 2). Telehealth use was weakly associated with sex, with females modestly more likely to use telehealth (OR 1.03, 95% CI 1.01–1.06, P < 0.0001).

People who reported not speaking English as a first language were 40% less likely to use telehealth (OR 0.61, 95% CI 0.58–0.62, P < 0.0001) than English speakers (Table 2 – Model 1). People born outside of Australia and who do not use English as a first language were 22% less likely to use telehealth (OR 0.77, 95% CI 0.75–0.79, P < 0.0001) (Table 2 – Model 1). In contrast, people born in English-speaking countries different from Australia did not differ from people born in Australia (OR 1.01, 95% CI 0.99–1.05, P = 0.32). Adjusted by country of birth, the multivariate model showed that people who reported not using English were 39% less likely to use telehealth (OR 0.62, 95% CI 0.59–0.65, P < 0.0001) (Table 2 – Model 2). People requiring interpreter services had 66% fewer occasions of telehealth services (OR 0.34, 95% CI 0.32–0.36, P < 0.0001) than those not requiring interpreter services (Table 2 – Model 2). This association did not vary after adjusting by their use of English or country of birth (OR 0.33, 95% CI 0.31–0.36, P < 0.0001). Among telehealth users, people requiring an interpreter were 13% less likely to use videoconference than phone (after adjusting for country of birth and use of English, OR 0.87, 95% CI 0.77–0.98, P < 0.0001) (Supplementary Table S2). People living in higher socioeconomic areas were more likely to access telehealth (OR 1.05, 95% CI 1.04–1.05, P < 0.0001) (Table 2) and videoconferencing (OR 1.06, 95% CI 1.05–1.07, P < 0.0001) (Supplementary Table S2). However, people in the lowest decile of IRSD were 50% less likely to use telehealth (OR 0.49, 95% CI 0.48–0.50, P < 0.0001) compared to other categories in the same index.

Strengths and limitations

A strength of our study is the large dataset (n = 153 427) from four different (geographically and socio-demographically) tertiary hospitals, which supports the transferability of our results.

Our study has important limitations worth noting. First, data from tertiary hospitals in Queensland may overrepresent telehealth use among patients with complex needs compared to the general population in other settings (e.g. primary care) or other Australian states. Fortunately, our results sufficiently expose the inequitable distribution of telehealth use among those who do not speak English as a first language or those living in low socioeconomic conditions. Second, this study is an observational snapshot in time and does not imply causation. Third, we assumed care was provided in English, the dominant language in Australia, but some informal translation could have occurred (e.g. bilingual staff or patients’ care or family members) and our data cannot identify nor adjust for any such instance of this nature.