Knowledge on Indoor Air Quality (K-IAQ): Development and Evaluation of a Questionnaire Through the Application of Item Response Theory

Abstract

Indoor air pollution is a major cause of noncommunicable diseases, and increasing people’s knowledge about the related risks is a key action for prevention. Many studies describe questionnaires for evaluating knowledge on indoor air quality that often involve selected population groups and take time to fill out. This study describes the validation of a questionnaire built “ad hoc” that aims to be easy to fill out, reliable, and valid. The validation process integrated two psychometric approaches: the Classical Test Theory (CTT), which uses the Kuder–Richardson 20 (KR-20) formula to measure the internal consistency and reliability of the questionnaire as a whole, and the Item Response Theory (IRT), which evaluates each statement (item)’s validity. The questionnaire, distributed using social media to a self-selected sample of people, reached a sample of 621 subjects. In terms of internal consistency, the questionnaire was found to be satisfactory, with a KR-20 value of 0.74 (CI 0.71–0.77). The IRT analysis showed that the statements included in the questionnaire can distinguish between high-performing and low-performing interviewees, since 100% of the items reached a value of the “discrimination parameter a_j” that was within or above the recommended range. In terms of difficulty, many statements (53.3%) showed a low level of difficulty, obtaining a low “difficulty parameter b_j” value, while another 20% of the items showed a high level of difficulty. Regarding the pseudo-guessing parameter, known as the c-parameter, the probability of answering correctly for a low-performing interviewee was observed in three items (1, 6, and 9), and the same statements fell outside the range for all three parameters evaluated in the IRT. The application of the IRT highlights the criticality of some questions that would not have emerged using the CTT approach alone. Although the questionnaire is acceptable overall, it will be appropriate to evaluate whether to revise or exclude the critical questions in order to improve the instrument’s performance.

1. Introduction

Air pollution is a major cause of noncommunicable diseases, including stroke, ischemic heart disease, chronic obstructive pulmonary disease (COPD), and lung cancer. [1]. According to the World Health Organization (WHO), indoor air pollution was responsible for approximately 3.2 million deaths globally in 2020 [1,2], including over 237,000 deaths of children under the age of 5 [2]. The combined effects of outdoor air pollution and indoor air pollution are estimated to cause approximately 6.7 million premature deaths each year [2].

For these reasons, the disease burden attributable to air pollution, overall, was in the top 5 out of 87 risk factors in the global assessment [3], alongside other major health risks such as an unhealthy diet and tobacco smoking [4].

The effects of indoor air pollution are closely related to the concentration of the pollutant and the exposure duration. Although, in closed environments (e.g., homes), the concentrations of some pollutants may be low, their overall contribution to exposure is relevant due to the time people spend in those places [4]. In fact, most people, in industrialized nations, spend 80–90% of their time indoors [1].

Given the health relevance of this topic, in order to prevent adverse effects on the population, the World Health Organization (WHO) and many countries have defined threshold values [5] for several pollutants of health interest.

However, having specific regulations and/or recommendations does not necessarily mean that these regulations will be strictly applied, for example due to misconceptions or cultural barriers. Empowering people can contribute to reducing the adverse effects of indoor pollution, but this requires awareness-raising and risk communication, which, in turn, involve increasing environmental awareness and promoting behaviors and practices able to reduce exposure [6,7,8,9]. Providing accessible information on indoor environmental exposures can also facilitate practical outcomes, such as the adoption of exposure-reducing behaviors [10] and the planning of appropriate mitigation measures [11].

In this field, the availability of reliable tools for evaluating a population’s knowledge needs can support the planning and realization of effective training programs and preventive actions [12].

In the literature, rather complex tools for measuring knowledge on indoor pollution are available, including investigations on knowledge alone [13,14], Knowledge, Attitude, and Behavior (KAB) [10,11,15,16], and knowledge, attitudes, and practices [12,17,18,19,20,21,22,23]. Some of them focus on target groups (e.g., students, pregnant women, hospitalized patients) [10,12,13,16,17,21,22,24], while others investigate the general population [15,18,19,20,23].

Depending on the study, different methodologies are adopted to reach the target population, such as online surveys [15,19,20,21,22], face-to-face interviews [10,11,13,17,18], and self-administered questionnaires in specific selected places [12]. The formulations of questions and answers are equally varied (e.g., true/false statements, multiple-choice questions, open-ended questions).

In terms of contents, some studies aim to evaluate knowledge about the sources of pollutants, health risks, and mitigation measures [11,16,21], focusing mainly on health hazards [13,15,20,21,23,24,25]. Other studies focus on the perception of indoor air quality and how this influences behavior [10,26,27], on environmental risks in general [12], or on chemical air pollution and the related risks [18,28].

Many of these studies concern contexts with habits, cultures, living conditions, and regulations that are very different from those in Italy. Therefore, the content of the questionnaires used in these surveys should be adapted to the specific needs of the country in which they are used.

To measure conceptual categories such as attitudes, interests, and skills, it is necessary to use objective and structured tools, based on statistical methodologies, connected to the definition of measurement scales. It is therefore necessary to define criteria to validate the different constructs in order to guarantee acceptable indices of reliability and validity for the test [12,29].

In this field, to validate tests, two main psychometric approaches can be used: the Classical Test Theory (CTT) [30,31,32] and the Item Response Theory (IRT) [29,31,32,33].

The CTT represents the most widespread approach to measuring latent traits, i.e., intrinsic unobservable characteristics, such as the degree of knowledge of a topic, emotional states, and clinical conditions, through subjective responses to single questions contained in specific questionnaires [34,35]. The element of interest is estimated by counting the number of responses provided by the person (overall score). Therefore, the entire questionnaire constitutes the unit of analysis for a given skill sought.

The CTT has been applied in different studies [31,36,37] as it is linked to the reliability and internal coherence of the questionnaire (e.g., through the calculation of Cronbach’s α, KR20, and KR21) [38,39] for dichotomous and/or polytomous variables. The difficulty level of the test is evaluated only by the percentage of correct answers provided by the interviewees, and the statistical approach used with the CTT method does not allow us to separate the attributes, associated with people, from the characteristics of the questions.

In contrast, the IRT [29,31,32] focuses the researcher’s attention not on the entire questionnaire, as in the CTT, but on the characteristics of each item, estimating numerical parameters related to it (such as the difficulty of the question, the level of discrimination, and the casuality of the response).

The use of the IRT approach allows us to both differentiate the subjects, based on the outcome of the evaluation test, and determine the relative difficulty of the questions included in the estimation scale. Therefore, the IRT bases the probability of success of the test on the following two axioms:

• Identifying the (personal) degree of a specific skill (e.g., numerical reasoning, sector knowledge) in answering targeted questions;
• Recognizing the threshold of difficulty of the task required (e.g., solving a technical problem).

In short, the IRT evaluates the ability to respond correctly for a given level of difficulty. From this, it follows that a person should have a greater chance of responding equally correctly to simpler questions [40].

Thanks to its characteristics, the IRT can be used to evaluate the psychometric properties of an existing scale and its items, but also to optimally shorten the scale when necessary and to evaluate the performance of the new scale. Therefore, IRT modeling—if well used—can support the realization of precise, valid, and relatively brief tools, resulting in a lower response burden and a higher rate of acceptance [29].

Given the above, the aim of this study was to evaluate a new questionnaire specially developed to understand a population’s knowledge about indoor air quality (K-IAQ). In particular, the investigation concerned the evaluation of the level of knowledge regarding the sources of indoor pollutants, the health hazards, and the risk mitigation actions. The validation process aimed to evaluate its reliability and ability to be understood in order to elicit consistent and correct answers.

2. Materials and Methods

2.1. The Questionnaire

The questionnaire was designed following a review of the existing literature on the tools available to measure knowledge and behaviors related to indoor air quality [11,12,13,14,15,16,17,18,19,20,21,22,25,26,27]. In light of the review, the new questionnaire mainly focused on assessing knowledge about some chemical, biological, and physical indoor hazards, their sources, and possible measures to mitigate exposure [2,5,6,41,42,43,44].

The questionnaire includes 30 true/false statements. Seven of these investigate knowledge about specific indoor health hazards (n.1, Indoor air can carry microorganisms that are dangerous for health; n.12, The development of mold in the house favors the appearance of allergies; n.13, Asbestos is a carcinogenic material; n.15, Household pesticides can induce headaches and nausea; n.17, Children exposed to secondhand smoke in the home are more at risk than other children to develop lung infections and asthma; n.19, The presence of a cat in the home can trigger an allergy in susceptible individuals; n.30, A household air temperature below 18° can cause respiratory diseases and mental health problems).

Another 11 statements refer to possible sources of pollutants in confined spaces (n.7, Mites multiply in wool mattresses and pillows; n.10, Cigarette smoke also emits radioactive substances; n.11, Radon gas is concentrated above all in cellars and floors next to the ground; n.14, Sound-absorbing panels can release dangerous substances; n.18, People lose heat and vapor in the environment; n.21, Gas stoves and cooking foods can emit toxic or irritating gases; n.23, Cigarette smoke that permeates fabrics and furnishings in homes can release formaldehyde and benzene; n.25, Formaldehyde emissions from furnishings tend to decrease rapidly; n.26, Carbon dioxide (CO₂) in indoor air is produced only by plants; n.27, Cosmetics can release irritating and/or toxic volatile organic compounds; n.28, Some stain removers can be toxic and pollutants).

The last 12 items concern possible mitigation strategies or risk-preventive actions (n.2, In winter it is not necessary to open the windows of the house; n.3, The kitchen fume extraction hoods must be connected to the respective flues; n.4, Radon gas is used to heat the home; n.5, A dehumidifier is enough to purify the air; n.6, The room’s ventilation serves to remove bad smells only; n.8, The airtight fixtures of the windows favor the exchange of air in the rooms; n.9, In the bathrooms, the air exchange reduces the accumulation of humidity from condensation; n.16, From the garages located under the buildings, the exhaust fumes of the vehicles can go back into the houses; n.20, For cleaning floors, it is always necessary to use a disinfectant; n.22, Filters of air conditioning systems must not be cleaned; n.24, The labels on the chemicals used in the home also provide information on their toxicity; n.29, The quality of indoor air depends only on the quality of the outdoor air).

Compared with other similar studies [16], the proposed questionnaire focuses on the risk factors commonly found in Italian indoor environments and also considers the regulations in effect in Italy.

After it was processed, a pool of indoor air quality experts with different backgrounds (chemists, biologists, doctors, architects, and engineers) revised the questionnaire to check its scientific soundness.

The questionnaire, in its final version, has been structured into two parts. The first part collects some personal information (gender, age, geographical area of residence) anonymously in order to guarantee the privacy of the interviewees and encourage sincere answers, reducing the fear of possible judgment. In this section, the number of inhabitants at their living place is also required in order to better understand the urban or rural context of the interviewees. These data should allow us to contextualize the answers provided and to subsequently carry out a more in-depth analysis of the results, considering various demographic and socio-economic factors.

The second part of the questionnaire includes 30 true/false statements regarding indoor air quality and its fundamental aspects.

2.2. The Sample Selection

To perform the study, a convenience sampling approach was chosen. The questionnaire was distributed using social media to a self-selected sample of people in order to achieve a good response rate in a quick and effective way. The target was to reach the critical threshold of at least 500 interviews [29,45,46,47] to give the results reliability for the validation process. The responses were automatically collected in a Google Form built “ad hoc”, which organized data in an easily analyzable format. In particular, the responses to the questionnaire were exported to an Excel file (.xlsx) and then successively exported to other specific pieces of software to be analyzed.

Although this sampling method has some limitations, such as possible selection bias [48,49], it was chosen for the reasons described previously and, above all, because our goal was not to have a representative sample of the population but rather to reach enough people to validate the tool. Indeed, it permitted low-cost dissemination of the questionnaire compared with other possible approaches (face-to-face interviews, telephone interviews, and email distribution) [50] and the achievement of the critical threshold of interviews [29,45,46,47] required to give reliability to the results in a relatively short time.

2.3. Statistical Analysis

To validate the questionnaire, the following two approaches were used: an evaluation of each statement (item)’s characteristics using the IRT and an evaluation of the internal consistency and reliability of the questionnaire using the KR-20 formula.

As already described, the IRT evaluates the characteristics of each statement (item) in terms of difficulty, level of discrimination, and causality of the response, putting therefore in relationship examinee responses to item characteristics and examinee ability to examinee answer.

In mathematical terms, the IRT model associates a probability value (p) of the response of the i-th subject to the j-th item as a function of the difference between the person parameter and the item parameter using a logistic approach. Therefore, taking into consideration the original |M| matrix (n × m, 30 items x > 500 people), we calculated the overall score for each column (item), for the N = 30 variables, with dichotomous coding (1 = correct answer; 0 = incorrect answer or no answer provided). The probability was evaluated using a logistic function described by the 3-parameter IRT model (3PL) [51,52] for the case of dichotomous variables (X_ij = 1|0). The probability of a positive response for the j-th item is given by:

$$p_{ij}=\mathrm{Pr}\left(X_{ij}=1|{\theta }_i; a_j; b_j; c_j\right)=c_j+(1-c_j)\frac{exp\left\{a_j\left({\theta }_i-b_j\right)\right\}}{1+exp\left\{a_j\left({\theta }_i-b_j\right)\right\}}$$

(1)

where θ is the ability sought in person i, a_j is the discrimination parameter, b_j is the difficulty parameter, and c_j is the randomness parameter [51,52].

In detail:

• a_j (discrimination parameter): This parameter expresses the ability of an item to discriminate between candidates with different abilities [53]. It highlights different interpersonal attitudes, regardless of gender, country of origin, city, etc., used to interact with the questions, and high numbers reflect high abilities. It is given by the ratio between the fraction of correct answers and the total number of questions. An a_j parameter value closer to 0.0 indicates the absence of discrimination between respondents with high and low abilities. A discrimination index close to 1.0 is optimal [54]. A discrimination index equal to or greater than 0.3 is considered highly discriminatory [54]. For acceptable values, taking into account the aims of this study, the range 0.3 < a_j < 1.0 [54] was adopted, even though wider intervals have been chosen in the literature [53,55,56], generally for questionnaires with different purposes.
• b_j (item difficulty parameter): This parameter defines how difficult it is to respond to a question. Items that receive nearly uniform correct or incorrect responses can be described in terms of low and high difficulty. High values intercept difficult items, while low thresholds intercept easy questions. It follows that there is a point on the scale (of abilities) in which the probability of answering correctly is equal to 50% [53,56,57,58]. The range of acceptable values was 0.30 < b_j < 0.70 [52,53,54,56].
• c_j (randomness parameter): This parameter shows the probability that an interviewee with low ability can answer an item correctly [53]. It is calculated as 1/k, where k is the number of options given; for dichotomous variables, k = 2 and c = 0.5 (i.e., the interviewee has a 50% probability of answering the question correctly, despite not having relative skills) [45]. The randomness parameter (cj), also known as the pseudo-guessing parameter, has a theoretical range of 0.0 ≤ c ≤ 1.0 [59], but, in practice, values above 0.35 are not considered acceptable, hence the range 0.0 ≤ c ≤0.35 is usually adopted when the 3-parameter logistic model is used [56]. Therefore, the accepted range was 0.0 < c_j < 0.35 [52,53,56,57].

Overall, in the literature, several studies present different ranges depending on the topic addressed. Given the relevance and specificity of the topic addressed in this study, a narrower acceptable range (a_j, b_j, c_j) was chosen to ensure greater accuracy to identify statements requiring improvements.

The calculation of the parameters was performed in a Microsoft Windows environment with the open-source software R and some calculation algorithms (MS Excel, v.2016) associating the overall score for each question with the probability of a correct answer (p_%). Items and people were therefore placed on the same scale (CAT), starting from the assumption that the questions are linearly independent of each other (local independence) for a given level of ability (θ) and that the characteristic trait, in this case knowledge, about the topic covered is a single trait (unidimensional latent trait), as the questions have the sole purpose of identifying only one specific characteristic.

The results obtained were processed to verify the overall reliability of the instrument and the achievement of the 3 described criteria for each item to identify the questions that documented the lowest level of difficulty, reliability, and chance that could affect the final result.

Subsequently, to estimate the internal consistency and reliability of the questionnaire, the Kuder–Richardson 20 (KR-20) formula was used, which is a specific case of Cronbach’s α computed for dichotomous scores [60,61,62]. It is an index of internal consistency that measures how closely a set of items correlate with each other.

The formula to calculate the KR-20 value for a test with K test items numbered i = 1 to K is

$$r=\frac{K}{K-1}\left[1-\frac{{\sum }_{i=1}^K{p}_i q_i}{{\sigma }_X^2}\right]$$

(2)

where:

p_i = is the proportion of correct responses to the test item;

i, q_i = is the proportion of incorrect responses to test item i (so that p_i + q_i = 1);

${\sigma }_X^2$ = is the variance;

The values of the KR-20 test can range from 0.00 to 1.00. Conventionally, a KR-20 value greater than 0.70 is considered acceptable.

3. Results

A total of 621 people answered the questionnaire.

Table 1. Description of the population sample.

Variable	Detail	N.	%
Gender *	Male	306	49.35
	Female	314	50.65
	Missing	1
Age (year) *	≤20 y	115	18.61
	21–30 y	266	43.04
	31–40 y	84	13.59
	41–50 y	44	7.12
	51–60 y	59	9.55
	>60 y	50	8.09
	missing	3
Origin*	North Italy	47	7.59
	Central Italy	385	62.20
	South Italy	112	18.09
	Other Country	75	12.12
	missing	2
City size (N. Inhabitants) *	<5000	122	19.71
	5000–50,000	193	31.18
	>50,000–500,000	84	13.57
	>500,000	221	35.70
	missing	1
Total		621	100

* The percentage calculation was performed on the respondents.

describes the characteristics of the population sample.

The gender distribution shows a slight predominance of female participants (314), equal to 50.65% of the interviewees.

The sample, with an average age of 32.5 years, includes subjects from 12 years to 84 years, with the highest percentage of participants in the age group 20–29 years (266 subjects, 43.04%).

About 88% (554 subjects) of the interviewees are of Italian nationality. Among them, the majority come from central regions (69.4%) or from southern ones (20.2%). Only 47 subjects (8,4% of Italian participants) come from North Italy. The 75 foreigners (12% of the total sample), all belonging to the age group ≤ 30 years, are university students from several different countries.

The interviewees come from cities of different sizes. In particular, the largest number of them live in large cities with more than 500,000 inhabitants (35.70%) and in intermediate-size cities with 5000 and 50,000 inhabitants (31.18%). Many of them (122 subjects, 19.71%) live in small inhabited centers (<5000 inhabitants), and the remaining 13.54% live in medium-sized cities (50,000–500,000 inhabitants).

The overall analysis of the answers to the 30 true/false statements shows that 63.95% of these were correct 63.95% (11,913 vs. 18,630). On average, participants obtained a score of 19.18 correct answers (median = 20), despite the wide variability in scores (from 0 to 30) at the individual level.

Figure 1 shows the frequency distribution of interviewees by scores reached. The scores were collected in categories, and the most represented category was that with the number of correct answers between 19 and 21. Overall, 366 respondents (equal to 58.93%) obtained a score ≥ 19. This highlights that, for the investigated population, the test was not difficult.

The characteristics of each statement were then evaluated using the IRT approach.

Table 2. Evaluation of the characteristics of each item using the IRT.

Question Number n.	Question	Correct Answer %	Discrimination Parameter aj	Item Difficulty Parameter bj	Randomness Parameter cj
			0.3 < aj < 1.0	0.30 < bj <0.70	0.0 < cj < 0.35
1	Indoor air can carry microorganisms that are dangerous for health	82.8	1.49	−0.64	0.5
2	In winter it is not necessary to open the windows of the house	77.9	0.97	−1.15	0.21
3	The kitchen fume extraction hoods must be connected to the respective flues	52.1	0.83	0.75	0.24
4	Radon gas is used to heat the home	47.5	1.11	0.53	0.14
5	A dehumidifier is enough to purify the air	83.8	0.86	−1.82	0.21
6	The room’s ventilation serves to remove bad smells only	92.4	1.04	−1.67	0.5
7	Mites multiply in wool mattresses and pillows	80.9	0.65	−1.94	0.2
8	The airtight fixtures of the windows favor the exchange of air in the rooms	51.4	0.87	0.52	0.18
9	In the bathrooms, the air exchange reduces the accumulation of humidity from condensation	86.2	1.11	−1.07	0.5
10	Cigarette smoke also emits radioactive substances	23.7	1.09	1.96	0.11
11	Radon gas is concentrated above all in cellars and floors next to the ground	45.9	1.83	0.6	0.18
12	The development of mold in the house favors the appearance of allergies	82.0	0.74	−1.82	0.22
13	Asbestos is a carcinogenic material	88.4	0.52	−3.54	0.22
14	Sound-absorbing panels can release dangerous substances	30.2	1.03	1.4	0.09
15	Household pesticides can induce headaches and nausea	79.5	1.05	−1.27	0.18
16	From the garages located under the buildings, the exhaust fumes of the vehicles can go back into the houses	59.3	1.16	0.34	0.29
17	Children exposed to secondhand smoke in the home are more at risk than other children to develop lung infections and asthma	90.8	0.75	−2.26	0.5
18	People lose heat and vapor in the environment	84.1	0.93	−1.76	0.2
19	The presence of a cat in the home can trigger an allergy in susceptible individuals	54.7	0.55	0.44	0.19
20	For cleaning floors, it is always necessary to use a disinfectant	34.5	0.91	1.94	0.2
21	Gas stoves and cooking foods can emit toxic or irritating gases	46.6	1.19	0.66	0.18
22	Filters of air conditioning systems must not be cleaned	54.8	0.53	0.38	0.18
23	Cigarette smoke that permeates fabrics and furnishings in homes can release formaldehyde and benzene	47.1	1.48	0.67	0.21
24	The labels on the chemicals used in the home also provide information on their toxicity	76.6	1.1	−1.08	0.16
25	Formaldehyde emissions from furnishings tend to decrease rapidly	26.2	2.24	1.31	0.13
26	Carbon dioxide (CO2) in indoor air is produced only by plants	81.9	1.11	−1.44	0.17
27	Cosmetics can release irritating and/or toxic volatile organic compounds	68.2	0.93	−0.55	0.19
28	Some stain removers can be toxic and pollutants	83.8	1.22	−1.48	0.18
29	The quality of indoor air depends only on the quality of the outdoor air	81.6	1.38	−1.24	0.15
30	A household air temperature below 18° can cause respiratory diseases and mental health problems	18.9	0.84	2.97	0.11

shows, for each statement, the percentage of correct answers and IRT results in terms of the difficulty of the question, discrimination ability, and casuality of the response.

The questionnaire distinguishes well individuals with low abilities from those with high abilities, since, in the column regarding the discrimination parameter (parameter a_j), many statements (16 out 30, 53.3%) show values over the recommended range and no other column shows values below the acceptability interval.

Regarding the difficulty of the statements (parameter b_j), 16 of them (53.3%) were easy, obtaining values below the recommended range. The other six statements (20.0%) showed a high level of difficulty, and the remaining eight questions were within the acceptability range.

A high percentage of statements (86.7%) fell within the recommended range for the randomness parameter c_j, and only four questions (13.3%) were filled in randomly.

Overall, using the IRT approach, three statements fell outside the range for all parameters (the discrimination parameter a_j, the difficulty parameter b_j, and the probability of random response parameter c_j), and they will require a careful evaluation.

The results were analyzed according to the three identified knowledge areas (sources of indoor pollutants, health hazards, and risk mitigation strategies).

With respect to health hazards, the only statement with unsatisfactory scores for all three parameters included in the IRT was number 1, regarding the potential spread of microorganisms dangerous to health through the air, which the majority of people answered correctly (82.8%). This item, despite having a very high discrimination capacity (a_j 1.49), revealed a very low level of difficulty (b_j −0.64) and a high probability of a random response (c_j 0.5), over the acceptable limit.

In the same area, items 12, 13, 15, and 17 were found to be easy (with b_j values always below the minimum limit) and have a high percentage of correct answers (82.0%, 88.4%, and 79.5%, respectively). Item number 30 was too difficult (b_j equal to 2.97). Questionnaire statement 19 (“The presence of a cat in the home can trigger an allergy in susceptible individuals”), however, was well-formulated, as the application of the IRT was within the accepted range for the discrimination parameter a_j, the item difficulty parameter b_j, and the probability of a random response parameter c_j.

In the area of sources of indoor pollutants, the greatest critical issues were found in the b_j difficulty parameter. Some items (numbers 7, 18, 26, 27, and 28) were easy, with negative b_j parameter values, while others (numbers 10, 14, and 25) were difficult, with b_j parameter values greater than 1. Eight out of eleven items (specifically numbers 10, 11, 14, 21, 23, 25, 26, and 28) showed a very high discrimination capacity, with a_j values always greater than 1.0 (the optimal value). No statements had all ITR parameters outside the recommended ranges.

Finally, two questions belonging to the “risk mitigation strategy” area (item number 6 and item number 9) require attention, showing a high discrimination value (1.04 and 1.11, respectively), but a low difficulty level (−1.67 and −1.07, respectively) and a high probability that an interviewee with low ability can answer the item correctly (cj value equal to 0.5).

In the same area, several items (numbers 4, 6, 9, 16, 24, and 29) showed a_j parameter values above the selected range; these values are still close to 1.0, which is considered the optimal value of discrimination. As regards the difficulty parameter, some items (2, 5, 24, and 29) were found to be easy, with a negative b_j value, and showed a percentage of correct answers higher than 75%. Still in the same context, when applying the IRT, statements 8 and 22 showed values that fell within the accepted ranges for the discrimination parameter a_j, the item difficulty parameter b_j and the probability of a random response parameter c_j.

Overall, the items within the questionnaire were found to correlate with each other, as shown by the evaluation of internal consistency and reliability with the application of the KR-20 formula, which reached a satisfactory value (0.74; CI 0.71–0.77).

4. Discussion

Assessing a population’s knowledge about indoor air quality is a key aspect because it can provide useful information for creating training programs accessible to that population, planning appropriate mitigation measures, and, at the same time, influencing people’s behaviors in order to reduce exposure.

In addition, it is fundamental to use questionnaires able to measure exactly what the researcher wants to measure. Most studies that evaluate knowledge, attitudes, and practices regarding indoor air quality used the CTT to analyze the internal consistency of the questionnaire (e.g., Cronbach’s α) [18,22,25]. In the present study, the CTT was used, which apples the KR-20 formula [62], a variant of Cronbach’s α calculated for dichotomous variables, obtaining a good correlation between items (0.74; CI 0.71–0.77).

Overall, this psychometric approach provides useful information for the validation of a questionnaire, but it does not allow for an exact evaluation of the intrinsic characteristics of each item and, therefore, for highlighting the major critical issues on which to intervene to improve the instrument.

For this reason, using the IRT, it was possible to relate the participant’s answers to the characteristics of the item and the interviewee’s ability to respond, highlighting some critical issues to review [29,31,32,46,63]. For example, although it was considered reliable (given the results of the KR-20 test), the questionnaire was overall quite easy for the participants, with 53.3% of statements obtaining a value of the b_j parameter lower than 0.3. This is also documented by the high average number of correct answers, equal to 19.18, and 58.93% of respondents with scores ≥ 19.

This good level of knowledge may of course depend on the actual easiness of the questions, but it may also depend on some characteristics of the participants, such as their education, sources of information, and previous experience. In the literature, educational level is also widely associated with a higher knowledge score [12,18]. In particular, the impact of prior experience on knowledge is an important factor, since it can influence knowledge independently of education [16]. Therefore, considering literacy as the only criterion for defining indoor environmental knowledge level may not be sufficient. Unfortunately, this study did not investigate these aspects. To better understand whether a high average score depends on education or on other factors, it will be necessary to go in depth on that in future studies, being a crucial point for the application of the tool.

It should also be noted that five items (nos. 10, 14, 20, 25, and 30) were difficult for respondents to understand, resulting in very low percentages of correct answers, always below 35.0%, and with values for the IRT b_j parameter far above the acceptable value.

Three of these statements (nos. 10, 14, and 25), which concern the sources of pollutants (no.10, Cigarette smoke also emits radioactive substances; no.14, Sound-absorbing panels can release dangerous substances; no.25, Formaldehyde emissions from furnishings tend to decrease rapidly), are all questions that imply a good level of scientific literacy [16]. The same applies to statement no.30 (“A household air temperature below 18 °C can cause respiratory diseases and mental health problems”), which refers to health hazards. Montuori et al. (2020), for example, highlight in their study that a high percentage of respondents knew that formaldehyde is a household chemical pollutant, but less than 50.0% of the population knew that formaldehyde is classified as a carcinogen [18].

This difficulty may be due to the fact that the questionnaire was administered to a panel of experts and the scale reflects environmental problems that are known and recognized by experts, but not necessarily by the general public. This difficulty highlights the need to more effectively transfer knowledge to the public to influence behavior and guide more effective preventive measures [16]. In particular, the low percentage of correct answers to question 10 (“Cigarette smoke also emits radioactive substances”) is concerning, because, despite numerous information campaigns on smoking-related health risks [26], the substances [26] cigarettes can emit are little-known, just as it is rarely known that vaping via electronic cigarettes and smoking “heat-not-burn” devices are also sources of air pollution [21].

Finally, question 20 (“For cleaning floors, it is always necessary to use a disinfectant”), which refers to a mitigation action aimed at environmental hygiene [12,16,24], is of particular interest because it concerns a commonly encountered form of misconduct, probably influenced by the media and advertising for disinfectant products.

The IRT results show that three other statements, one regarding the health hazard area (n.1, Indoor air can carry microorganisms that are dangerous for health) and the other two regarding mitigation measures (n.6, The room’s ventilation serves to remove bad smells only; n.9, In the bathrooms, the air exchange reduces the accumulation of humidity from condensation), had all three parameters outside the considered intervals, although the percentage of correct answers to these items was high. This is probably due to the good discrimination of statements and the low degree of difficulty of the answer, factors that could have generated a random answer to these questions due to insufficient attention or little commitment to answering. However, most people did not respond randomly to the items, except for four statements (nos. 1, 6, 9, and 17) resulting in random responses from the interviewees.

However, it should not be overlooked that this study does not precisely quantify the average knowledge of the population, as it was created with other objectives. Therefore, this study shows some limitations. For example, it does not allow us to draw generalizable conclusions, as the results obtained are based on a convenience sample, which can lead to low representativeness due to selection bias. Regarding this point, it is important to emphasize that this study was intended only to obtain a sufficient sample size to validate the instrument and individual questions and not to draw definitive conclusions regarding the level of knowledge of the general population. This last issue will be the subject of other studies to be carried out in the future.

Another limitation is the recruitment of participants via social media. According to some authors [49,64], population samples obtained in this way tend to be biased and rarely accurately represent the general population, as social media users may have different demographic or behavioral characteristics [48]. Furthermore, they may provide inaccurate responses, resulting in poor-quality data. These issues can lead to a higher rate of error in estimates, as highlighted by some previous studies [49,64].

Not all authors agree on these critical issues, preferring these methods to established ones [50]. Certainly, web-based survey methods are advantageous in terms of the response speed, costs, and response rate, proving highly efficient and effective for conducting preliminary studies such as the one we conducted.

5. Conclusions

Overall, given its internal consistency, the questionnaire demonstrates a good ability to measure the level of knowledge about indoor air quality.

However, considering that the IRT highlighted the difficulty of some questions, and that these questions primarily required a certain background in scientific subjects, these questions may be unsuitable for measuring IAQ knowledge among the general population. In a certain sense, this aspect could lead to a different formulation of the question, focusing it more on practical aspects [16]. At the same time, given the health importance of IAQ, identifying these knowledge gaps while leaving the questions unchanged could help design more targeted and effective information campaigns.

In conclusion, we believe that the developed measurement scale represents a good starting point that could be further improved in future studies, including the introduction of questions designed to understand factors related to respondents’ varying levels of knowledge about indoor air quality.