Readability and beyond - Health literacy and numeracy and COVID-19 communications in early childhood education: Are we communicating effectively?

Objective Analyse the linguistic and numerical complexity of COVID-19-related health information communicated from Australian national and state governments and health agencies to national and local early childhood education (ECE) settings. Methods Publicly available health information (n = 630) was collected from Australian national and state governments and health agencies, and ECE agencies and service providers. A purposive sample of documents (n = 33) from 2020 to 2021 was analysed inductively and deductively combining readability, health numeracy and linguistic analyses and focusing on the most frequent actionable health advice topics. Results COVID-19 health advice most frequently related to hygiene, distancing and exclusion. Readability scores in 79% (n = 23) of documents were above the recommended grade 6 reading level for the public. Advice was delivered using direct linguistic strategies (n = 288), indirect strategies (n = 73), and frequent mitigating hedges (n = 142). Most numerical concepts were relatively simple, but lacked elaborative features (e.g., analogies) and/or required subjective interpretation. Conclusion COVID-19 health advice available to the ECE sector included linguistic and numerical information open to mis/interpretation making it difficult to understand and implement. Practice Implications Combining readability scores with measures of linguistic and numerical complexity offers a more holistic approach to assessing accessibility of health advice and improving health literacy among its recipients.


Introduction
From 2020, the challenge of governments communicating accessible and rapidly changing information about the COVID-19 pandemic to its citizens became a global concern [1,2]. In the absence of an Australian Centre for Disease Control and Prevention (CDC), the responsibility for providing medical and public health information nationally sits with the federal government's Department of Health [3]. However, the unique geography and federated system of governance and healthcare delivery means state and local governments also have a mandate to provide health directives. All governments used a combination of lockdowns, isolation and exclusion, hygiene measures, physical distancing, and testing requirements to contain the spread of COVID-19 in the absence of vaccines, in 2020. Information to the public came mostly from the national and state departments of health and was delivered through government websites [4], and news channels (radio, television, web-platforms).
During a time of rapidly changing public health orders, early childhood education (ECE) providers were positioned as an essential workforce, with services remaining open, despite periods of strict lock down of workplaces, for much of the duration of the pandemic. As an essential workforce [5,6], the ECE sector was required to implement and convey complex and frequently changing health messages.
As part of their standard professional practice, early childhood educators are required to adopt pro-active measures to protect the safety, health and wellbeing of children, their families and others in their workplace. These include public health protection measures such as implementing health and hygiene practices that minimise the transmission of diseases [7]. During the pandemic, the ECE sector was tasked with expanding their health practices to include up-to-date, evidence-informed interventions to prevent the spread of COVID-19. This study investigated the complexity of health advice in the ECE sector during the first year of the pandemic.

Health literacyreadability measures
Crucial to an effective pandemic response, which requires individuals to change their behaviour, is the effective communication of public health information. To respond quickly to changing public health messages people use their health literacy skills for accessing, processing, understanding and acting on the messages. Health literacy refers to "an individual's ability to find, understand and use health information in order to promote and maintain health" [8] and is an important psychosocial determinant of health outcomes [9].
Readability is a key facilitator for understanding health information [10][11][12][13]. Health professionals recommend materials for the general public, particularly in emergencies, be readable at a grade 6 reading level to have maximum impact [14]. Approximately 14% of Australians have a reading level at or below grade 6 level, and 44% fall below a grade 11/12 level [15,16] -the latter being required for broad participation in work, education, training, and society [15,17]. Government guidelines recommend writing at a grade 7 level to ensure information is accessible to most Australians [16]. Australian early childhood educators are diversely qualified including Certificate III, Diploma and Bachelor's degree or above [18]. Additionally, English may not be the primary language of families or members of the ECE workforce. Educators often work with families that may have lower literacy levels.
Readability scores are often used to assess accessibility of health information [10][11][12], but in isolation they cannot give definite answers as to how easily information can be accessed, processed, understood and applied. Applied linguistics investigates real world language use providing further insights into how health information can be parsed and applied. Linguistic and health literacy research into communication of health information during COVID-19 has focused amongst other topics on the infodemic facing the general public [19,20], issues for multilingual communities [21,22], metaphors in COVID-19 discourse [23] and discourse features of health advice (i.e., how is advice phrased) that may shape if and how advice is consumed by the target audiences [19,[24][25][26][27][28][29]. Emerging linguistic analyses of how COVID-19 health advice was communicated world-wide via websites [27], social media [25] and in documents [19,26,28,30], have crucial implications for the effectiveness of public health communication [31,32]. Linguistic analyses have not yet investigated communication of health advice involving information brokers at the intersection of health and ECE sectors.

Health numeracy
Another important but often overlooked element of health literacy is health numeracy. An individual's health numeracy is "the capacity to access, process, interpret, communicate, and act on numerical, quantitative, graphical, biostatistical, and probabilistic health information needed to make effective health decisions" [33]. Numbers feature in many reports about the spread of the virus (e.g., number of confirmed cases in the context of number of people tested, hospitalisation and mortality rates, R [reproduction number], exponential growth, and flattening the curve [34]).
Providing easy to comprehend numerical information is crucial as numeracy represents a specific challenge for many Australians [35]. The Australian Council for Educational Research [36] recommended numeracy be made a priority target for upskilling, with specific mention to supporting women and educators. Approximately 20% of adults are performing at or below level 1 [17], equivalent to grade 6 and below [8], where competencies are limited to carrying out one-step or simple processes such as counting or arithmetic with whole numbers in common, concrete contexts with explicit mathematical content and little or no text or distractors [37]. Many of the tasks required for interpretation of numerical information related to COVID-19 are assumed to fall under levels 2 and 3 [38], equivalent to grade 7-12 ability levels.
Considering health literacy and numeracy demands placed on individuals to interpret COVID-19 health information and advice, this study aimed to analyse accessibility of material provided by information brokers working to communicate health advice from Australian national and state governments and health agencies to national and local ECE settings. Combining readability, health numeracy and linguistic analyses, we focussed on the most frequently actionable [39] COVID-19 health advice topics (that moved audience to action) found in documents produced by health and education authorities, for variable audiences including Culturally and Linguistically Diverse (CALD) and Aboriginal and Torres Strait Islander (ATSI) communities who speak a diversity of community languages other than English.

Methods
The current data form one part of a mixed-methods collaborative study [40] to investigate how the ECE sector experienced communication of health information during the first year of the COVID-19 pandemic in Australia. Partner organisation included health (Western Sydney Local Health District), union (United Workers Union) and ECE organisations (Community Connections Solutions Australia, Community Early Learning Australia, Early Childhood Australia, Early Learning and Care Council of Australia, Family Day Care Australia, KU Children's Service, The Front Project). Here we report on the readability, linguistic and numerical analysis of publicly available documents communicating COVID-19 related health information to diverse audiences. Ethics approval was not required for this analysis of publicly available secondary data.

Data collection
From February 2021 to April 2021, LS and MRD (with research assistance) sourced publicly available documents by systematically searching the websites of Australian national and state health authorities and education departments, work health safety organisations (e.g., SafeWork Australia), and peak ECE and partner organisations for COVID-19 related health information documents. Documents were included if they contained COVID-19 related health information and were produced between March 2020 and April 2021. Relevant documents linked within sourced documents were also collected and crossreferences recorded. If necessary, previous versions of linked documents were sourced using internet archives (e.g., Wayback machine).
All sourced documents were saved locally, logged in an Excel spreadsheet with a unique ID recording information on: producing organisation services (e.g. ECE, Health, Education, and ATSI and CALD organisations), produced at national/state/territory level, document title, type (e.g. fact sheet, poster, newsletter, email), publication channel (e.g. website, social media ([Facebook, YouTube]), target audience (families/children, ECE frontline staff, ECE organisation/managers), date, version number, language, linked translation, cross references, and researcher comments and quality impressions.

Data selection
Overall, we identified 630 publicly available documents. From these, a purposive sample of documents was selected for detailed analysis based on a selection framework developed and refined by the team (See Table 1). The selection was guided by achieving a balance between three producing services (ECE/education, Health, ATSI/CALD) and three target audiences (families/children, ECE Frontline staff, ECE organisations/managers). The initial target of 45 documents for detailed analysis included 15 documents per producing service and 5 documents per audience and sector (see Table 1). As some cells did not yield sufficient documents, we only included a total of 33 documents in the analysis. The selection followed a two-step process: 1) key documents were selected as i) indicated as crucial by partner organisations, ii) indicated high distribution (high number of cross-references) and iii) researcher quality impression (excellent and/or poor); 2) purposive selection was made to attempt to achieve even coverage across i) document types, ii) Australian states, iii) producing organisation, and iv) publication time.

Data analysis
In an iterative, data narrowing, multi-step process, all 33 documents were analysed inductively and deductively by LS, MRD, RB and DT.
Step 1 was conducted using a readability scoring website, and steps 2-4 used NVIVO (version 12, QSR International).
In Step 1, readability of the whole document (or the first 3000 words) was assessed for each written document (n = 29) with four transcripts of videos excluded from analysis. Scoring was completed using a website [41] that provides scores for the most commonly used readability measures. Documents were cleaned and prepared for analysis following guidance in Ferguson et al. [42] and Stossel [12]. Numerical score (e.g., 44.2) and graded score (e.g., "difficult to read") for each document were recorded for Flesch Reading Ease score, Gunning Fog, Flesch-Kincaid Grade Level, SMOG Index, and the Readability Consensus (grade level, reading level, reader's age). Using the readability consensus scores, documents were then assigned into one of three categories: 1) at or below 6th grade reading level ((very) easy to read), 2) at 7th-8th grade reading level (standard/average), or 3) at 9th or above grade reading level ((very) difficult/fairly difficult to read).
In Step 2, LS and MRD inductively identified all instances and types of health information (e.g., hygiene, physical distancing, symptoms and disease, exclusion, testing, etc.). In Step 3, LS and MRD deductively mapped only the sections of documents coded as instances of health information onto the six core features of the World Health Organisation's (WHO) Strategic Communications Framework for effective communications [39]. The six features are: accessible, actionable, credible and trusted, relevant, timely, and understandable. Following discussions, the research team decided the fourth and final step of the analysis should focus on the three most frequent types of health information data coded under the WHO Actionable code which target behaviour changes and moving readers to action [39].
In Step 4a, MRD and LS deductively coded the linguistic strategies ( Table 2) employed to communicate actionable health information/ advice [25,43,44]. Codes were assigned for one major linguistic strategy identifying overall level of directness (from direct to indirect) and any relevant number of additional mitigating strategies (e.g., hedges such as 'consider x'). In Step 4b, RB and DT deductively analysed the complexity of numerical information in the purposively sampled documents. Several conceptual frameworks have been developed to reflect the difficulty of numerical concepts found in math curricula and assessments [33,45]. We followed the framework proposed by Joram et al. [46], which was developed in the context of these previous frameworks and the Principles and Standards for School Mathematics [47]. Each piece of numerical information was coded as type of mathematical concept (see Table 3) and use of elaborative features (e.g., pictures, diagrams, anecdotes, and metaphors that may help the reader connect less familiar to more familiar content). If a concept had multiple components, e.g., 20s, it was coded into multiple categories, in this case both as a whole number (20) and as measurement (seconds). Concepts in the coding scheme are organised to reflect their respective difficulty [46,47]. DT (a research and practitioner expert in mathematics education) also aligned mathematical concepts to their grade of introduction in the Australian mathematics curriculum.
Elaborative features were coded as any element that represented the numerical concept in an alternate way (e.g., diagram), elaborated the concept verbally through explanations, specifications, or examples, or provided explicit support for the reader's comprehension/memory in the form of a mnemonic or benchmark for a quantitative concept.

Readability scores
Overall, the average word count for documents was 1508 words (w).  "If an officer feels unwell for any reason, they will not be conducting visits and your service will be called and advised." Probability "Before I have a test, I might feel lots of different emotions." Direct Declarative "Sharing of food should be actively monitored and discouraged." Imperative "Clean your hands for 20 seconds with soap and water or an alcohol-based hand sanitizer." Negative imperative "Don't share food or drink." Performative "[…] the NSW Government has advised that electronic check-in methods for all businesses in NSW are now mandatory." Mitigating Strategies External modifiers Positive outcome *Statement that expresses the positive effect of following the advice "We are doing great progress to control the virus." Reason or explanation *Statement that explains the reason or cause that justifies the provided advice "Following these measures will help you protect yourself and others." Statement of empathy *Statement that expresses the speaker's empathy towards the hearer "I understand how you must be feeling." Internal modifiers -Downgraders Hedges *Mitigating word or construction used to lessen the impact of an utterance/ reduce the force of another word or phrase. "You are making kind of a statement with the pants though." "Yes, I mean it might be but it still seems to me at the moment that perhaps it's not a good idea." Politeness markers *Words of phrases that express politeness "Please." Internal modifiers -Upgraders Intensifiers *A word, especially an adverb or adjective, that has little meaning itself but is used to add force to another adjective, verb, or adverb: "very", "so really" Word counts for documents aimed at families and children (1,038w) were shorter than those for frontline staff (1,567w) or ECE organisations/managers (1,990w). Documents produced by ECE/Education (1,095w) were shorter than those by the ATSI/CALD (2,284w) or Health services (1,498w). Table 4 provides an overview of readability scores across target audience and producing services. Of all written documents, 79% (n = 23) had a readability score above the grade 6 reading level recommended for online (health) information [42]. Readability scores above the suggested level were observed among all services, with between 80% (ATSI/CALD) and 88% (ECE) of documents per service being above grade 6 level.
Only 25% of documents aimed at families and children were above standard year 7-8 readability levels. No documents for ECE frontline staff were produced at an easy to read level. The great majority of documents (91%) targeted at ECE organisations/managers were difficult to read, one with a reading level of university graduates.
Evaluating word count and readability scores, documents aimed at families and children were short and more than 75% readable at an average grade 7-8 or very/easy to read level. In contrast, documents for ECE frontline staff and ECE organisations/managers were lengthy and largely (50%) or predominately (91%) difficult to read.

Types of health information and WHO strategic communications framework
We identified 13 types of health information (See Table 5). Across all documents, hygiene, symptoms and disease, distancing, and exclusion were most frequently mentioned. The highest percentage of information related to symptoms and disease, and mental health were found in documents aimed at Families (64% and 58%) and the highest percentage for lines of Communication and operations were found in documents targeting ECE organisations/managers (59% and 67%). While differences in content focus might be expected, we also found that certain aspects of health information were underrepresented for certain audiences. For example, frontline workers received the lowest percentage of information related to mental health (16%) and ECE organisations received the lowest percentage of information related to the virus and how its spreads (4%). For in-depth linguistic and numerical analysis, we selected hygiene, distancing, and exclusion as the three most frequent types of health information because of their more balanced distribution across all audiences.
Within the WHO strategic framework [39], Understandable, Actionable, and Relevant were the most frequent coded aspects of health information, followed by Credible and Trusted, Timely and Accessible (see Table 5).

Linguistic strategies for actionable health information
As summarised in Table 6, across the three most frequent types of health information, most actionable health guidance, i.e., implementable advice, was delivered using direct strategies (total n = 288: imperative n = 161, declarative n = 112), followed by indirect hinting strategies (n = 43) and conventionally indirect strategies (n = 30; conditional n = 20). For external modifiers, providing reasons and/or explanations was used almost exclusively (n = 39; hygiene n = 29, distancing n = 5, exclusion n = 5), with positive outcomes used once and statement of empathy not at all. For example, reasons or explanations were predominantly given for health information relating to hygiene (n = 29) but spread evenly among audiences: […] to help prevent sickness [10053-targeting Families]. So you won't catch diseases [10033 -targeting Children].
[…] to kill the virus and protect yourself and others [10060 -targeting Frontline].
[…] to help minimise the risk of exposure to COVID-19 for education and care service employees [10077-targeting EC organisations].
For internal modifiers, hedges were most frequent (total n = 142: distancing n = 65, exclusion n = 33, hygiene n = 44), followed by intensifiers (n = 39). Intensifiers such as adjectives (always, important, highly, immediately, critical) or verbs (will, must) were observed across all audiences and most frequently related to hygiene (n = 18), followed by exclusion (n = 15) and distancing (n = 6): Staff must always wash hands with soap and water for 20 s or use a hand sanitiser before and after performing routine care [10060 -Frontline]. It is critical that any staff member or child who becomes unwell while at a service returns home [10727 -EC organisation]. Always comply with request from the driver. [10250 -Families].
The phrasing and the large amounts of hedges used across all linguistic strategies left implementation of actionable advice open to interpretation (see Table 7). Lexical hedges (e.g., should, can/could, may/might, (re)consider, try, avoid, (dis)encourage, little bit where(ever)  possible, unless necessary, when appropriate) and syntactical hedges (e.g., conditionals if…) were also found in direct communication strategies such as imperatives and declaratives. For example, instead of using a clear unhedged imperative "run an indoor/outdoor program", actionable health information phrases often contained hedged phrases such as "consider running an indoor/outdoor program, wherever possible" [10060]. Also, the use of unhedged and hedged advice within the same document further complicated how audiences may interpret or implement advice such as the clear imperative "Don't share food or drink." compared to a mitigated declaration "Sharing of food should be actively monitored and discouraged" [10060]. Most hedges occurred in documents that were above recommended reading levels (76%, n = 100) as compared to those at grade levels 7-8 (21%, n = 29), or at grade 6 or below (2%, n = 3) (Readability scores were not obtained for four video transcript documents).

Numerical analysis
Analysis of the 33 documents examined the frequency of each type of numerical concept for documents targeted at each audience. Numerical concepts were all at the basic and intermediate level, with most frequent use of whole numbers and measurement (see Table 8). Elaborative features in the form of benchmark or analogy was seen eight times in four separate documents; one of these documents also included an accompanying picture. For example, benchmarks such as "we should be able to spin around without touching anyone" [10002] with an accompanying picture of a child with outstretched arms in a spinning motion described physical distancing in a document targeted at families. Another document targeted at ECE organisations described ways in which educators could talk to children about physical distancing, including the analogy of "pretending you have a bubble around you", and the benchmark of "keeping an arm's length apart" [10082]. Two documents provided by state government departments targeted at families provided the benchmark of "the length of two Happy Birthday songs" [10240,10410] as the recommended 20 s for handwashing (similar elaboration of hand washing time was seen in all four documents).
Two documents used pictures only (no additional text) with the numerical information, e.g., a picture of two children standing with arms outstretched with clear distance between their hands, and an indication of 1.5 metres with a double headed arrow, was used in one document targeted at families to depict physical distancing [10002]. A similar depiction was seen in website media from one large service provider to families [10333], describing the services health and safety plan (two people, depicted by icons, separated by a double headed arrow labelled as 1.5 m).
Secondly, analysis considered the types of numerical concept for  each health behaviour (see Table 9). Intermediate level numerical concepts were more frequent for hygiene behaviours (usually in relation to cleaning requirements/solutions) and for distancing behaviours. Numerical concepts were rare in communications regarding exclusion. Table 9 provides examples in each category. Some documents did not present specific numerical information that fell within the coding scheme, but did require an understanding of area (Grade 2) and ratios (Grade 7) to ensure physical distancing while also complying with required staff:child ratios.
The following examples were in a website communication from a state Department of Education targeted at frontline staff [10060]: Consider having less children at each table and use more tables to allow more space between children. Wherever possible (e.g., weather dependent) and where you have enough staffing for adequate supervision, consider operating an indoor/outdoor program for the full day/session. This naturally provides for more space for the children and the setup of more activities for children to engage in.
There was also frequent use of words that refer to rate but are nonspecific, e.g., regularly, frequently.

Discussion
Our study takes an important first step towards innovating how comprehensibility and accessibility of written health information and advice is assessed. We emphasise the need to move beyond readability as a sole measure and we seek to advance the field of public health communication by extending the existing paradigm to also include detailed analysis of linguistic and numerical complexity in assessing the accessibility of health advice. Drawing on the diverse strengths of our interdisciplinary team in applied linguistics, numeracy and public heath during the interpretation of the data, we highlight how multiple perspectives can provide a more in-depth understanding and evidence for the improvement of COVID-19 health advice.
Our study identified actionable COVID-19 health advice in publicly available documents most frequently related to hygiene, distancing and exclusion. Much of the content of health communication information provided to, and created by, the ECE service was '(very) difficult to read' as it was written for those with higher than grade 6 reading level. Documentation aimed at Frontline staff and ECE organisations/managers were consistently written at higher than recommended reading levels. Educators' demographic characteristics may play an important role, with international research suggesting that more qualified and experienced educators have a greater sense of control over one's health, and feel more prepared to deal with pandemic health outbreaks than their lesser qualified and experienced counterparts [48,49]. These feelings, alongside variable (health) literacy and training levels [50] may also impact their confidence in reading and implementing health advice, especially if they are (very) difficult to read.
Frequent use of hedges in information on health-related actions meant that implementation of advice was open to mis/interpretation. This finding adds to the growing bodies of evidence that, frequently, reading level standards are not being met, with online COVID-related information on large numbers of websites exceeding 6th grade level readability [51,52] and even information designated as "easy read" exceeding a grade 8 reading level [53]. Information from government  and public health bodies tends to achieve better readability scores, more in line with recommended levels [11].
In contrast, much of the numerical information communicated for the targeted behaviours was relatively simple, with none requiring advanced numerical skills such as computation or comprehension of risk and probability. However, there were many instances where nonspecific references to quantity were used, which, as with the linguistic information, left the communication open to mis/interpretation. There was also minimal use of elaborative features; the provision of highly explicit numerical information (to reduce the need to make inferences) and the use of elaborative features are both recommendations for effective presentation of numerical information in health media [54].
While we purposively sampled documents according to the producing services and targeted audiences, we did not specifically select documents that were high in numerical content. The actionable items analysed here had minimal demands in terms of numerical comprehension. However, other COVID-19 related information is much more numerically complex, e.g., when assessing the risk of catching the disease: "even when we had higher rates of disease in [state], only 1.5% of cases were in school students (8 cases) and one case was in a teacher, which were mostly related to travel, all of whom had mild illness and have fully recovered" [10367, factsheet from state Department of Health to frontline staff] Interpersonal pragmaticsstudying how language use within its socio-cultural contexts shapes meaning-making and relationship-and trust-building [32,55] contributes a new measure, rather than relying on readability as a sole marker of accessibility and understanding health advice. Our findings support previous research on COVID-19 health advice [27,29] with documents most frequently featuring direct advice structures (imperatives and declaratives) which should make advice as clear and explicit as possible [25]. Reasons for advice can mitigate perceived burden [25] and were frequently provided for hygiene but not for distancing or exclusion. Adding to the evidence base on actionable COVID-19 health advice, we have demonstrated that the ease of implementation may be thwarted by the amount of mitigating language and hedges used. Hedges occurred most often in documents above recommended reading levels, making them "difficult to read" and interpret. While internal modifiers including politeness markers and hedges may be intended to soften directives and enhance engagement with the audience [25,28,56,57], these discourse features can lead to direct advice becoming less clear and explicit.

Conclusion
In examining what information brokers in ECE settings were facing when communicating and receiving health information across a range of national, state and local producing services, our analysis showed that health advice, while phrased as clear directives, could be hard to understand and implement due to linguistic complexity, or by having both linguistic and numerical information that is open to mis/interpretation.
We have demonstrated that including linguistic and health numeracy analysis in addition to simple measures of readability allows for a more comprehensive investigation of written health advice. Moving beyond only readability as a marker of accessibility showed that health advice can be complex to interpret and implement.

Practice implications
In Australia today, the ECE sector has the potential to deliver health information to more than 700,000 families and 200,000 staff. If the health brokering potential of the sector is to be realised, however, the information that the sector receives must be clear and easily understandable. Readability, numerical and linguistic complexity affect the effectiveness of health information produced by and for ECE services. Our study has practical implications advocating for a pluralistic approach that incorporates readability scores as well as measures of linguistic and numerical complexity offering a more holistic approach to assess and improve accessibility of health advice related to COVID-19, other crises or general and disease-specific health information. To achieve production of materials at appropriate reading levels (at grade level 6 [42]), building awareness among, and upskilling content producers in the concept of readability would assist in better communicating critical health information to a diverse service, with a mix of literacy and numeracy capabilities. Rethinking the use of hedges and overall clarity of numerical concepts in health advice is critical for content producers to consider in improving public health messaging. This would include, but not be limited to, increasing levels of directness and consistently

Declaration of Competing Interest
All articles must include a separate file containing a statement declaring any competing interests that relate to any authors. Competing interests are defined as those potential influences that may undermine the objectivity, integrity, or perceived conflict of interest of a publication. Interactions that occur within 5 years before submission date of an article are pertinent.