Psychometric testing of the teacher food and nutrition-related health and wellbeing questionnaire

Abstract

Background

Current evaluation of teacher wellbeing rarely includes assessment of diet quality or influential food and nutrition (FN) constructs. Diet quality, and FN constructs such as food skills, are increasingly recognised across health research for their associations with mental health, stress regulation, and overall wellbeing, making their inclusion essential in future iterations of teacher wellbeing assessment.

Method

This study evaluates the reliability and construct validity of the teacher food and nutrition-related health wellbeing questionnaire (TFNQ) created using the online survey platform QuestionPro to measure FN constructs alongside teacher wellbeing outcomes. An intra-class correlation coefficient (ICC) ≥ 0.50 was used to establish test-retest reliability, using matched sample data with ≤ 17 days between responses, from both timepoints (dataset one and two, n = 99). Construct validity was assessed with data set one (timepoint 1, n = 438) using confirmatory factor analysis with global fit criteria (root mean square error of approximation ≤ 0.08, comparative fit index ≥ 0.90 and standardised root mean residual ≤ 0.05).

Results

Five hundred and twenty-two primary and secondary schoolteachers from across Australia completed the TFNQ at two timepoints. Six of the seven TFNQ sub-scales achieved an ICC ≥ 0.70; the remaining sub-scale, eating social norms at school achieved an ICC of 0.46. All sub-scales except social support, met at least two of the three global fit criteria for construct validity.

Conclusions

The TFNQ is the first purpose-built evaluation tool to incorporate a set of valid and reliable sub-scales and single-item measures. The TFNQ will facilitate research that aims to explore the contribution of teacher FN practices with teacher health and wellbeing and inform the development of practical lifestyle focused solutions.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12889-026-26329-0.

Background

Teachers experience high levels of stress and burnout, which can be further impacted by poor food and nutrition (FN) practices and vice versa [1]. As FN role models, teachers are important influencers and health promoters, yet little is known about teachers’ personal FN practices and potential relationships with wellbeing indicators. Of the teachers identified within the 2020/2021 Australian National Health Survey, only 8.7% meet vegetable intake guidelines, with 45.8% and 25.5% meeting fruit and physical activity guidelines respectively [2], placing teacher dietary patterns in line with the general population [3]. Diet quality influences health-related outcomes [4], with research now linking diet quality along with other FN constructs (i.e., a scale or group of questions that measure an aspect of food or nutrition) such as cooking confidence, with mental health and wellbeing outcomes [5]. Wellness interventions that include diet quality and culinary skills as a strategy to manage stress [6] have emerged across research literature.

While evidence supports the role of FN constructs in relation to wellbeing, existing teacher wellbeing measures largely ignore diet-related influences on brain health and psychological wellbeing [7–9], excluding FN factors and diet quality from evaluation. To this end, validated tools to measure FN constructs as a component of teacher wellbeing are minimal [10] with a general lack of evaluation of teacher wellbeing programs offered [11].

Beyond health and wellbeing of teachers, the World Health Organization (WHO), and United Nations Educational Scientific and Cultural Organization (UNESCO) recognise the importance of teachers as health promoters, with the role of teacher wellbeing noted as critical in being effective educators and health promoters [7, 12]. Teachers act as role models and health promoters for their students, colleagues and the broader school community with important implications for whole of school health promotion [12]. The WHO and UNESCO recognise the additional responsibilities of teachers as health promoters and role models of FN practices reinforcing the need for ongoing teacher training to be positive food, nutrition and health influencers within schools [12]. Teachers provided with teaching-focused FN education, report increased self-efficacy in delivering FN lesson material [13, 14]. Furthermore, school health promotion interventions that incorporate teacher support improve teacher uptake and program outcomes [15].

While globally FN constructs have been explored in teachers, the range of constructs examined, how they have been defined and the types of questions used to measure each, have been diverse and inconsistent [10]. There is limited data on teacher FN practices, including level of food skills [10, 16] and the influence of schools’ eating social norms on teacher dietary intake [17]. This limits overall understanding of FN issues specific to teachers and is an impediment to development of informed, inclusive wellbeing education to support teachers in maintaining healthy FN practices. Despite evidence linking diet quality and FN-related behaviours to mental health and stress regulation, most teacher wellbeing measures omit these constructs, leaving a critical gap in both understanding and the design of effective interventions [10]. The teacher food and nutrition-related health and wellbeing questionnaire (TFNQ) was designed to provide a systematic, evidence-based approach to evaluating the contribution of FN constructs to teacher wellbeing and teacher self-efficacy as FN influencers within schools [18]. Improving teacher wellbeing is vital because teachers shape the health culture of schools for both collegues and students; however, existing wellbeing tools rarely assess FN constructs, limiting opportunities for targeted support [10, 12].

The TFNQ was designed to create a screening and evaluation tool to specifically investigate the relationships between FN constructs, such as food skills confidence [16] and eating social norms at school [17], alongside teacher wellbeing outcomes, including burnout [19], stress, and coping [20]. This was done in two phases. In phase one, published elsewhere, TFNQ priority constructs were identified from the literature, or where gaps existed, they were created de novo [10]. Constructs consistent with the Determinants of Nutrition and Eating (DONE) Framework, were identified with specific interest in constructs that were potentially modifiable and had capacity for impact at the population level selected [21, 22]. The influence of the school environment and culture on teacher’s FN practices encouraged the creation of the social support sub-scale and inclusion of the eating social norms at school sub-scale [17, 23]. Bespoke measures were created to enhance investigation of teacher eating behaviours and levels of professional confidence as FN influencers including the meal frequency, home food responsibilities and professional FN confidence sub-scales [18].

Phase two used a co-consultation approach with a two-round online Delphi methodology to establish content validation of the TFNQ, with feedback from 23 national and international experts from education, nutrition behaviour, dietetics, and psychology [18]. Content validation was established for the TFNQ, resulting in the inclusion of eight sub-scales and 21 single-item measures, representing 26 food, nutrition, wellbeing, and demographic screening constructs and three lifestyle covariates [18]. While the details of the Delphi study are published elsewhere [18], Table 1 in this article provides a summary of the sub-scales and single-item measures included.

Table 1.

Overview of all sub-scales and single-item measures within the preliminary TFNQ

FAVVA (Fruit And Vegetable VAriety) Index, TFNQ (Teacher Food and Nutrition-related health and wellbeing Questionnaire)

The current paper summarises the psychometric testing conducted to evaluate the TFNQ measures within a teacher population. The study aim was to establish test-retest reliability of sub-scales and single-item measures and construct validity of sub-scales within the TFNQ to enable evaluation of FN as a component of health and wellbeing in teachers.

Methods

Procedure

To assess the reliability of the measures within the TFNQ a test-retest methodology was used with the TFNQ distributed to teachers at two timepoints, approximately two weeks apart. The first data set from questionnaire one was further used to conduct construct validity using confirmatory factors analysis (CFA) of the three existing and four newly developed sub-scales within the TFNQ.

Participants

Primary, and secondary teachers across Australia aged ≥ 18 years, with internet access and currently employed in an Australian school were recruited. No exclusions were set for year level or learning area (subject taught), with the aim of recruiting a diverse sample. Social media channels were used to distributed study information between January and February of 2023 (including Facebook, Twitter and LinkedIn) using a rolling recruitment process to complete the first TFNQ. QuestionPro survey software was used to distribute the TFNQ (QuestionPro Inc, Austin Tx) at both timepoints. Recruitment material included a direct link to the first questionnaire, with participant email collected for distribution of the second questionnaire at timepoint-two. For the second questionnaire three reminder emails were sent out to participants with incomplete responses only, with a two-week period provided for completion of questionnaire two. Screening questions asked participants to identify the school they were currently teaching at. Implied consent to participate was assumed by completion of the first questionnaire, which was estimated to take approximately 30 minutes per timepoint. Participant time was acknowledged with the issue of a $30 GiftPay voucher if they completed the TFNQ at both timepoints.

Sample size was calculated to allow adequate numbers to conduct construct analysis using the CFA minimum recommendations for goodness of fit assessment ratio of cases to free parameters method or N: q of 10:1 to 20:1 [24–26]. The TFNQ uses predominately single factor models of between five to seven indicators and 15–21 free parameters, with the food skills confidence [16] sub-scale, the largest in the TFNQ, having 19 indicators and 50 + parameters. A sample size of 420–500 was considered desirable to strike a balance between the practicality of recruiting working teachers, cost and statistical power. For test-retest a minimum sample of approximately n = 30 repeat participants at timepoint two was required to sufficiently assess reliability using the intra-class correlation coefficient (ICC) method selected [27, 28].

Instrumentation

The TFNQ was developed using an expert Delphi process. Table 1 provides an overview of the included sub-scales and single-item measures, while Appendix 1 provides a detailed breakdown of the questions and scoring procedures.

Data analysis

Data cleaning procedures included de-identification and assignment of unique participant codes to ensure confidentiality and traceability. Prior to analysis records from non-Australian participants, duplicate responses (i.e., identified by repeated email addresses) and fast completions (i.e., indicative of insufficient engagement) were excluded to maintain data integrity.

Seven of the eight sub-scales, except for the Fruit And Vegetable VAriety (FAVVA) Index [29], and 21 single-item measures (excluding demographic screening questions) were tested to ensure normality principles applied to both data sets from timepoint one and timepoint two prior to reliability and validity testing. Skewness (≤ 2) and kurtosis (≤ 4) were used as the basis for normality testing due to the size of the data sets and variety of sub-scales and single-item measures included within the TFNQ [30, 31].

Reliability testing

All reliability analyses were conducted using IBM SPSS Statistics (Version 28). Test-retest reliability was established using ICC with a benchmark of ≥ 0.70 used to establish high reliability [28, 32]. In recognition of the diversity of included measures and variability among them in relation to temporal flux [33, 34], a lower benchmark of ≥ 0.50 was also used [27] to establish reliability and stability of rank order. Percentage mean difference was calculated to assess stability of the sample characteristics and for use as a secondary measure to provide additional evidence on the reliability of measures with an ICC value close to established benchmark [28, 32].

Test-retest reliability was to be completed using all matched data of all available participants, however an issue with the distribution of the questionnaire a timepoint two meant some participants received the second survey more than the planned 14-days. The sample of participants who completed the second survey in ≤ 17 days between responses (n = 99, responses range between 14 and 17 days) were therefore used to conduct test-retest analysis. This enabled analysis closer to the fortnight planned within the original project proposal with consideration of the wellbeing measures that were more susceptible to higher temporal flux [34]. The remaining participants (n = 279; responses ≥ 17 days) were included as a separate analysis to examine if the measures employed were displaying temporal variability.

Construct validity of sub-scales

Construct validity was conducted using R-Studio Statistics Software (RStudio Team, 2020) and the LAAVAN package [35] to assess the ability of the items within each sub-scale to provide an accurate measure of the construct being asssesed. Construct validity was assessed on three existing sub-scales from international populations (food skills confidence [16], burnout [19], eating social norms at school [17]) and four newly developed sub-scales (professional FN confidence, meal frequency, home food responsibilities, and social support) using CFA. To achieve construct validity, sub-scales were required to meet at least two of the three global fit criteria of a root mean square error of approximation (RMSEA) value of ≤ 0.08, a standardised root mean residual (SRMR) value of ≤ 0.05, and a comparative fit index (CFI) value ≥ 0.90. Additionally, local fit of each sub-scale was reviewed with factor loadings ≥ 0.60 preferred [24–26].

The FAVVA Index is a dietary screener previously assessed for comparative validity using adjusted regression analysis with mean food and nutrient intakes from the Australian Eating Survey Food Frequency Questionnaire and plasma carotenoid concentrations within an Australian adult population [29, 36]. With fruit and vegetable intake previously reported as a good indicator of positive wellbeing outcomes [37], the FAVVA index [29] was selected for inclusion in the TFNQ to reduce participant completion burden and provide a valid measure of fruit and vegetable intake and variety. As the FAVVA index was previously tested for comparative validity in a sample of Australians adults [29] it was excluded from the psychometric testing in this study.

Results

Figure 1 provides an overview of responses to questionnaire 1 (data set 1) and questionnaire 2 (data set 2) and their equivalent data set, with a summary of participant demographics at both timepoints presented in Table 2. Data set one included 56.8% males (n = 249), with most participants aged 31–45 (n = 399), 86.3% living with a partner and children between the ages of 0–20 years of age. Ninety-eight percent identified their ancestry as Australian (n = 433), with 15% acknowledging Aboriginal ancestry and 5.3% acknowledging Torres Strait Islander ancestry. The majority (n = 433) were currently on full-time contracts, with 81.3% of participants identifying their primary role as a classroom teacher (n = 356).

Fig. 1.

Questionnaire distribution

Table 2.

Participant demographics

Variable	Data set one N = 438 (%)	Data set two N = 480 (%)	Matched data ≤ 17days N = 99 (%)
Gender
Male	249 (56.8)	263 (54.8)	63 (63.6)
Female	189 (43.2)	217 (45.2)	36 (36.4)
Age in Years
<=30	12 (2.7)	9 (1.9)	2 (2.0)
31–45	399 (91.1)	441 (91.9)	95 (96.0)
46–55	27 (6.2)	30 (6.3)	2 (2.0)
56–65	0 (0)	0 (0)	0 (0)
Living Arrangement
Live with partner	38 (8.7)	8 (1.7)	2 (2.0)
Live with partner and children (0–20 years old)	378 (86.3)	451 (94.0)	95 (96.0)
Live with partner and grown children (Older than 20 years)	16 (3.7)	19 (4.0)	1 (1.0)
Single parent with young children (0–20 years old)	2 (0.5)	0 (0)	0 (0)
Live with other adults (shared accommodation)	2 (0.5)	1 (0.2)	0 (0)
Live with parent(s)	1 (0.2)	0 (0)	1 (1.0)
Live with parents and siblings	1 (0.2)	1 (0.2)	0 (0)
Ancestry
Australian	433 (98.9)	480 (100)	99 (100.0)
English	4 (0.9)	0 (0)	0 (0)
Scottish	1 (0.2)	0 (0)	0 (0)
Identify as having Aboriginal or Torres Strait Islander Ancestry
Yes, Aboriginal	68 (15.5)	106 (22.1)	34 (34.3)
Yes, Torres Strait Islander	23 (5.3)	44 (9.2)	11 (11.0)
Yes, Aboriginal and Torres Strait Islander	6 (1.4)	4 (0.8)	0 (0)
No	340 (77.6)	325 (67.7)	53 (53.5)
Prefer not to answer	1 (0.2)	1 (0.2)	1 (1.0)
Contract Type
Full time	433 (98.9)	478 (99.6)	98 (99.0)
Part-time ≤ 3 days per week	3 (0.7)	0 (0)	1 (1.0)
Part time > 3 days per week	1 (0.2)	2 (0.4)	0 (0)
Casual	1 (0.2)	0 (0)	0 (0)
Primary Role
Classroom Teacher	356 (81.3)	430 (89.6)	94 (94.9)
Head Teacher	65 (14.8)	43 (9.0)	5 (5.1)
Deputy Principal	10 (2.3)	6 (1.3)	0 (0)
School Principal	2 (0.5)	1 (0.2)	0 (0)
Casual teacher	1 (0.2)	0 (0)	0 (0)
Office or support staff	3 (0.7)	0 (0)	0 (0)
Other	1 (0.2)	0 (0)	0 (0)
Current School Sector
Catholic	24 (5.5)	22 (4.6)	3 (3.0)
Government	341 (77.9)	382 (79.6)	88 (88.9)
Independent	73 (16.7)	76 (15.8)	8 (8.1)
Level taught
Primary	119 (27.2)	136 (28.3)	61 (61.6)
Secondary	273 (62.3)	311 (64.8)	14 (14.1)
Across both primary and secondary	46 (10.5)	33 (6.9)	24 (24.2)
Learning Area
Primary teacher	38 (8.7)	50 (10.4)	16 (16.2)
Mathematics	78 (17.8)	88 (18.3)	28 (28.3)
English	62 (14.2)	65 (13.5)	20 (20.2)
Sciences	58 (13.2)	57 (11.9)	15 (15.2)
Humanities and Social Sciences	32 (7.3)	31 (6.5)	7 (7.1)
History	36 (8.2)	50 (10.4)	9 (9.1)
Health and Physical Education	103 (23.5)	102 (21.3)	3 (3.0)
Technologies	13 (3.0)	17 (3.5)	1 (1.0)
Languages	13 (3.0)	16 (3.3)	0 (0)
The Arts	4 (0.9)	4 (0.8)	0 (0)
Other	1 (0.2)	0 (0.0)	0 (0)

All sub-scales and single-item measures from data set one (n = 438) and data set two (n = 480) were normally distributed with a summary of test-retest reliability and construct validity results presented across Tables 3 and 4.

Table 3.

Reliability and construct validity results of all sub-scales

Sub-scale	Test-retest reliability				Construct validity
	ICC(r) (95% CI)	% Mean difference	ICC(r) (95% CI)	% Mean difference	χ2(df)	RMSEA [90% CI]	CFI	SRMR	Parameters
	≤ 17 days between responses (n = 99)		> 17 days between responses (n = 279)		Data set one: n = 438
Food Skills Confidence (n = 19)	0.87 (0.81–0.91)	3.57	0.59 (0.45–0.69)	−8.02	0.00(152)	0.08 [0.07–0.08]	0.68	0.07	57
Eating Social Norms at School (n = 7)	0.46 (0.19–0.64)	−0.07	0.22 (0.02–0.38)	3.48	0.00 (14)	0.15 [0.13–0.17]	0.75	0.07	21
Burnout (n = 4)	0.87 (0.80–0.91)	2.48	0.46 (0.31–0.57)	−10.92	0.034(2)	0.07 [0.02–0.14]	0.99	0.02	12
Professional FN Confidence (n = 4)	0.88 (0.80–0.92)	−3.33	0.11 (−0.13−0.30) NS	−0.55	0.00(2)	0.10 [0.05–0.16]	0.86	0.04	12
Meal Frequency (n = 3)	0.88 (0.88 − 0.82)	5.12	0.55 (0.42–0.65)	8.36	~(0)	0.00 [0.00–0.00]	1.00	0.00	9
Home Food Responsibilities (n = 5)	0.87 (0.81–91.81)	0.95	0.49 (0.35–0.60)	7.55	0.00(5)	0.11 [0.08–0.15]	0.91	0.05	15
Social Support (n = 4)	0.87 (0.80–91)	1.98	0.01 (−0.15−0.29) NS	−0.13	0.00(2)	0.34 [0.28–0.39]	0.75	0.11	12

Reliability tests used matched data from data set one and two, with construct validity results displayed in the table completed in data set one; NS = Not significant

Table 4.

Reliability results of all single-item measures

Single-item measure	Test-retest reliability
	ICC(r) (95% CI)	% Mean difference	ICC(r) (95% CI)	% Mean difference
	≤ 17 days between responses (n = 99)		> 17 days between responses (n = 279)
Wellbeing
Personal wellbeing
• Satisfied	0.70 (0.55–80.55)	6.95	−0.28 (−0.62- −0.02) NS	4.17
• Happy	0.67 (0.50–0.78)	8.96	0.43 (0.28–0.55)	−1.63
• Worthwhile	0.78 (0.67–0.85)	4.28	0.37 (0.21–0.51)	−2.32
• Anxious	0.64 (0.46–0.75)	8.90	0.47 (0.30–0.59)	27.18
Stress	0.86 (0.79–0.90)	2.33	0.23 (0.03–0.40)	0.15
Coping	0.42 (0.14–0.61)	5.49	0.43 (0.27–0.55)	−1.45
Perception of Physical Health	0.85 (0.76–0.90)	5.01	−0.16 (−0.47−0.09) NS	0.00
Perception of Mental Health	0.49 (0.21–0.67)	10.06	0.23 (0.03–0.38)	−10.75
Lifestyle covariates
Sleep Quality	0.62 (0.44–0.75)	−4.41	−0.47 (−0.86- −0.16) NS	−1.24
Sleep Quantity	0.24 (−0.12−0.49)	−4.57	0.53- (0.40–0.63)	2.29
Time in Physical Activity	0.80 (0.71–0.87)	4.51	0.24 (0.04–0.40)	−2.41
Days of Physical Activity	0.79 (0.63–0.87)	13.17	0.12 (−0.11−0.30) NS	−6.37
Alcohol Frequency	0.88 (0.83–0.92)	2.28	0.14 (−0.08−0.32) NS	16.39
Personal food and nutrition
Snack Frequency	0.47 (0.22–0.65)	−4.62	0.06 (−0.19−0.26) NS	−8.84
Take-away Frequency	0.74 (0.60–0.82)	−5.82	0.59 (0.44–0.69)	−15.22
Perception of Diet Quality	0.87 (0.80–0.91)	−0.27	−0.39 (−0.76- −0.10) NS	1.78
Perception Diet Importance for Physical Health	0.85 (0.78–0.90)	−2.01	−0.01 (−0.28−20) NS	2.56
Perception of Diet Importance for Mental Health	0.54 (0.32–0.69)	−1.79	0.35 (0.17–0.49)	−8.84
Meal Sharing Home	0.41 (0.12–0.60)	−11.76	0.16 (−0.06−0.34) NS	5.23
Meal Sharing School	0.47 (0.22–0.65)	−3.10	0.41 (0.25–0.54)	−10.63
Professional food and nutrition
Food Reward Frequency	0.71 (0.56–0.81)	9.58	0.25 (0.06–0.40)	−9.00

Reliability tests used matched data from data set one and two; NS = Not significant

Test-retest reliability

For measures taken ≤ 17 days apart, six of the seven sub-scales achieved an ICC ≥ 0.70 indicating high reliability. The sub-scale for eating social norms at school [17] had an ICC of 0.46, in line with the lower reliability benchmark of 0.50 indicative of slightly lower rank order repeatability. The low percentage mean difference of −0.07, indicated potentially limited systematic bias, within the longer time frames demonstrated in the sample responses with > 17 days. Therefore, the eating social norms sub-scale was accepted at this stage as a reliable measure over time. At this stage of testing as per established criteria, all seven of sub-scales were identified as reliable. Twelve of the 21 single-item measures achieved an ICC greater than or close to 0.70, meaning they had high reliability over time. Of the remaining nine single-item measures, six achieved an ICC greater than or close to the lower benchmark of 0.50. The remaining three single-item measures of coping [20], sleep quantity [38], and meal sharing at home had ICC values of 0.42, 0.24 and 0.41 respectively.

Tables 3 and 4 include test-retest analysis of matched data with ≤ 17 days between responses, and analysis with all matched pairs that had > 17 days between responses. As anticipated a decrease in rank order repeatability was observed for responses > 17 days. However, many of the measures maintained a percentage mean difference within a 5–10% window.

Construct validity

Of the three existing sub-scales tested within this study for an Australian population, food skills confidence [16], eating social norms at school [17] and burnout [19], only burnout achieved a suitable fit against all three global fit criteria (RMSEA ≤ 0.08, CFI ≥ 0.90 and SRMR ≤ 0.05), with food skills confidence [16] and eating social norms at school [17] sub-scales meeting only one of the three criteria. A review of the CFA factor loadings (latent variables) identified that of 19 items in the food skills confidence sub-scale [16], loadings ranged from 0.30 to 0.53, with eating social norms at school [17] (n = 7 items) ranging from 0.43 to 0.73, and burnout [19] (n = 4 items) 0.62–0.77 within the current analyses. Of the four newly developed sub-scales of home food responsibilities (n = 5 items), meal frequency (n = 3 items), professional FN confidence (n = 4 items), and social support (n = 4 items), all except the social support sub-scale obtained a SRMR value ≤ 0.05 and CFI ≥ 0.9, meeting two of the three global fit criteria set, with meal frequency being the only one to meet all three criteria, with factor loadings ranging from 0.69 to 0.92. Both the food skills confidence [16] and eating social norms at school [17] sub-scales had percentage mean differences ≤ 5%, with both achieving an ICC close to or greater than 0.50. However, both had poor CFA results in data set one. Of note, both constructs contained between 21 and 57 parameters respectively, indicating that a larger sample size may have been required to achieve adequate power to for CFA analysis [24–26]. When re-analysed in data set two, which had a larger sample size of (n = 480), the global fit values improved, with food skills confidence [16] meeting all three global fit criteria and hence construct validity (RMSEA = 0.05, [CI = 0.05–0.06], CFI = 0.93, SRMR = 0.04), with factor loadings ranging from 0.49 to 0.66. While eating social norms at school [17] met two of three global fit criteria with a CFI ≥ 0.90 and SRMR ≤ 0.05, and factor loading ranging from 0.50 to 0.73.

The revised TFNQ

The sub-scale for social support was flagged for further development, as it achieved reliability but failed to meet global fit criteria for construct validity. The reliability of the four social support questions were tested as both single-item measures and two summed scores. This was done by combining the two school social support items (e.g., school culture and colleagues) and two home social support items (e.g., family and friends) separately as alternative measures of home and school social support (Table 5). The single-item measures for school culture and family, and the summed scores for home and school social support, achieved ICCs ≥ 0.70 with the single-item measures for colleagues and friends each achieving an ICC close to the alternate benchmark of 0.50. The single-items for school culture and family, along with the two summed scores remained in the final TFNQ, with the two single-item measures for colleagues and friends marked for use as screening questions only (Table 5).

Table 5.

Test-retest analysis for social support single-item measures and summed scores

Single-item measure/ Summed score	Test-retest reliability
	ICC(r) (95% CI)	% Mean difference	ICC(r) (95% CI)	% Mean difference
	≤ 17 days between responses (n = 99)		> 17 days between responses (n = 279)
School culture	0.85 (0.78–0.90)	2.25	0.13 (−0.10−0.31) NS	2.81
Colleagues	0.48 (0.23–0.65)	−2.79	0.52 (0.39–0.63)	−4.90
Family	0.88 (0.80–0.91)	−1.64	0.08 (−0.16−0.27) NS	3.08
Friends	0.45 (0.19–0.63)	−6.04	0.51 (0.38–0.61)	−1.87
School social support summed score	0.76 (0.64–0.84)	−0.14	0.18 (−0.04−0.35)	−1.06
Home social support summed score	0.81 (0.71–0.87)	−3.70	−0.03 (−0.31−0.19) NS	0.65

Reliability tests used matched data from data set one and two; NS = Not significant

All sub-scales meet test-retest reliability as measures over time, with all except social support achieving adequate construct validity, meeting at least two of the three pre-determined global fit criteria. The proposed wellbeing single-item measure for personal wellbeing (anxious) [39] had an ICC closer to the lower benchmark of 0.50, at ICC = 0.64, with previous studies assessing measures of anxiety and wellbeing setting a Pearson’s Correlation coefficient of ≥ 0.50 as an acceptable reliability criterion [40]. Wellbeing measures can be more sensitive to situational changes with a higher temporal flux compared with measures used to assess more stable constructs such as reading comprehension over time [34]. Given this item held previous reliability and validity, it was kept in the revised TFNQ. Appendix 2 provides a summary of the changes made to the TFNQ following review of the reliability and construct analyses. Similarly, the single item for coping [20] achieved a low ICC = 0.42, which is comparable to the original study in which validity and reliability was established using Pearson’s correlation coefficient (r) to test temporal consistency, achieving r= 0.48 [20]. The five single-items for sleep quantity, meal sharing at home, meal sharing at school, frequency of snacking and mental health perception were flagged for alternate use as a screening questions or future review and re-testing in a different data set as these achieved low ICC and a review of percentage mean differences indicated that they are likely unsuitable as measures to track change over time. Table 6 provides a complete description of the differences between summed scores, single-item measures, sub-scales and screening questions as they relate to the TFNQ. Additionally, Table 7 provides an overview of all sub-scales, single-item measures, and screening questions in the revised TFNQ following completion of psychometric testing.

Table 6.

Definitions and description of questions and scale terminology used in the development of the TFNQ

Construct	A group of questions, or scales within the current study allocated within the current study to measure an aspect of food, nutrition, health, or wellbeing, i.e., personal wellbeing.
Sub-scale	A scale within the TFNQ made up of a collection of questions created to assess a construct i.e., burnout, food skills confidence with confirmed reliability and construct validity.
Single-item measure	A single, standalone question created to assess an aspect of food, nutrition, health or wellbeing, i.e., sleep quality, with confirmed reliability.
Summed score	A sum of two or more questions grouped within a construct, i.e., social support tested for reliability within the current study to measure home and school specific social support.
Screening question	Demographic or teacher characteristic single-item questions not tested for reliability as a measure over time, used to gather descriptive statistics. Additional questions that did not obtain suitable reliability as measures overtime were also kept as possible screening questions for used in descriptive analysis only.

Table 7.

Overview of all sub-scales and single-item measures within the revised TFNQ

Screening questions: Questions re-allocated to be used as screening questions for descriptive analysis data collection only along with other demographic and teacher characteristics questions, ‘n’ number of items/questions within a sub-scale or summed score

SS Summed Score, FAVVA Fruit And Vegetable VAriety Index

Discussion

The revised TFNQ contains seven valid and reliable sub-scales to assess burnout, meal frequency, home food responsibilities, food skills confidence, eating social norms at school, professional FN confidence, and fruit and vegetable intake and variety in the FAVVA Index. The sub-scale of social support was removed and replaced with two separate single-item measures with stronger reliability for school culture and family support along with two summed scores for home and school social support. The TFNQ can therefore be used as a tool in facilitating pre-intervention screening, program evaluation or adapted for use at a school-level to guide planning and development of staff-focused wellbeing education. The TFNQ offers a standardised method of assessing FN constructs including food skills confidence and wellbeing constructs, enabling investigation of/research efforts into the influence of FN constructs on teacher wellbeing. The current study established that seven sub-scales, and 16 of the 21 single-item measures tested meet reliability criteria, with six sub-scales meeting two or more of three global fit criteria for construct validity, except for social support. Social support was re-tested to establish two separate single-item measures of school culture and family social support, and two additional summed scores for school and home social support identified as reliable.

Contribution to public health promotion

Researchers are encouraged to review the contents of the TFNQ and select the range of constructs that best match study outcomes. A selection of single-item measures have been included to enable researchers a more parsimonious approach to screening food, nutrition and wellbeing constructs in time poor teachers, reducing both participant burden and analysis costs [33]. The TFNQ enables researchers and program developers the opportunity to pre-assessment food, nutrition and wellbeing constructs to guide planning of targeted education that aims to improve diet quality for improved teacher wellbeing outcomes [41]. School leaders could use the TFNQ to guide annual wellbeing program development or facilitate pre-assessment that includes a review of school-relevant constructs on teacher FN wellbeing including the influence of eating social norms at school and/or social support. With the two social support single-item measures and summed scores and eating social norms at school sub-scale providing a reliable approach to evaluate the influence of school culture and teacher home supports potentially impacting their ability to maintain healthy eating practices.

Limitations

This study recruited a convenience sample and self-reported data. While sample size achieved was suitable for proposed analysis, the final sample size was below recommended minimum of 500 participants needed to conduct robust CFA for the food skills confidence sub-scale. Additionally, despite asking participants which school they currently taught at using an Australian school database, some participants may have inaccurately reported their current employment status. This provided a learning for future research stages and is included as a recommendation to other researchers working with teachers to ensure participants use a professional, school email address to verify teaching status and school of current employment. In conducting test-retest reliability a fortnight between responses was planned within the current protocol. However, due to an issue with the distribution of the questionnaire at the second timepoint, the length between the first questionnaire and participant completion of the second was for most participants more than the initial fortnight proposed. Therefore, acknowledging that the TFNQ included measures sensitive to higher temporal flux [34], reliability testing was conducted using a smaller sample of matched pairs where participant responses were set at ≤ 17 days (n = 99).

This study focused on examining reliability and construct validity of the core sub-scales within the TFNQ. These psychometric tests were selected to establish reliability and validity for an Australian population of the existing sub-scales with previous psychometric testing in international populations. Additionally new sub-scales were selected for construct validity assessment following the establishment of content validity within a previous expert informed Delphi study to ensure the sub-scales suitably measured the constructs they were created to assess [18].

The TFNQ was not tested as a single measure with an overall total score, as the questionnaire focuses on individual food, nutrition and wellbeing constructs, each assessed as separate sub-scales or single-item measures. Criterion validity was not conducted as the focus was on FN constructs such as the food skills confidence sub-scale or the eating social norms at school sub-scale. While measures exist to assess food agency [42], food literacy [43], and nutrition knowledge [44] none were deemed suitably similar to conduct a valid criterion testing with included sub-scales like food skills confidence and eating social norms at school sub-scales.

FN is an emerging area of research with great diversity in the types of measures available. Therefore, this study successfully established test-retest reliability and construct validity using CFA of the individual components within the TFNQ to enable quality data collection and further research within this area, with a specific interest in the role of FN, including school specific measures such as eating social norms at school.

Conclusion

The current study introduces the TFNQ as screening and evaluation tool to track data over time. The TFNQ includes a collection of reliable and valid sub-scales and single-item measures that researchers and education bodies can use to inform the development of FN inclusive teacher health and wellbeing education. The TFNQ is a tool that measures FN constructs like food skills confidence and diet quality alongside measures of wellbeing such as personal wellbeing, stress, and coping. The TFNQ can be used to further research into how FN practices may be related to teacher wellbeing and their relationship with change in wellbeing outcomes over time. The TFNQ will improve intervention pre-screening and program evaluation methods to aid development of targeted education and support for teachers. Further research aims to use TFNQ as a validated tool to compile a normative database on the FN status of Australian teachers to inform current wellbeing program offerings while testing its application as an evaluation tool in a teacher FN-related health and wellbeing intervention.

Abbreviations

FN

Food and Nutrition

TFNQ

Teacher Food and Nutrition-related health and wellbeing Questionnaire

DONE Framework

Determinants of Nutrition and Eating Framework

FAVVA Index

Fruit And Vegetable VAriety Index

RMSEA

Root Mean Square Error of Approximation

CFI

Comparative Fit Index

SRMR

Standardised Root Mean Residual

CFA

Confirmatory Factor Analysis

CI

Confidence Interval

Authors’ contributions

TJ developed the testing and reliability protocol used in consultation with AM, CEC, VAS, and TB. Data analysis was conducted by TJ with assistance and guidance from AM. Manuscript development was completed by TJ with review and consultations contributions from AM, CEC, VAS, and TB.

Funding

This project is currently funded by the Teachers’ Health Foundation (This funding body had no role in the design or conduct of the study). TJ was supported by a Higher Degree by Research (HDR) scholarship Research Training Program (RTP) Stipend and RTP Allowance at the University of Newcastle and a King and Amy O’Malley Trust Postgraduate Scholarship with a PhD top-up scholarship from Teachers’ Health Foundation. VAS is supported by funding from the Hunter Medical Research Institute (HMRI). CEC is supported by an NHMRC Research Leadership Fellowship (L3, APP2009340).

Data availability

All data generated or analysed during this study are included in this published article [and its supplementary information files].

Declarations

Ethics approval and consent to participate

This study was conducted according to the guidelines laid down in the Declaration of Helsinki and all procedures involving the testing and reliability research study participants were approved by the University of Newcastle, College of Health, Medicine and Wellbeing Research Ethics Advisory Panel, Approval No. H-2022-0375. Written informed consent through survey completion (implied consent) was obtained from all subjects/patients.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.