The internet is a ubiquitous source of information. In 2017, 3.6 billion people worldwide have used the internet [1]. More than half of them (55%‑85%) used the internet to search for health-related information [2-5]. Seeking health information on the internet has many benefits, such as convenience, greater perceived anonymity regarding sensitive issues [6,7], and health service user empowerment [2,8-10]. However, a variety of downsides have also been described. The quality of many health-related websites has come under scrutiny [9,11-13]. The dangers of using search engines to find health information, eg, escalating searches that lead the user from researching harmless symptoms to self-diagnosing with life-threatening diseases, have been pointed out [12-16]. The users’ ability to appraise the material and filter irrelevant and unreliable information has also been questioned [11,12,16-19]. Some negative consequences of the relationship between health providers and patients have been reported [20,21]. Examples for such consequences comprise patients requesting inappropriate treatments on the basis of inaccurate, or inaccurately understood, Web-based information and frustration on the health professional’s side (due to challenges to their expert opinion and time used to explain) and the patient’s side (due to perceived unresponsiveness of the health professional to the Web-based researches).

People do not just search for information regarding their own health, but they also search for information relevant to the health of others in their family, particularly their children [3,4]. The reported prevalence of such searches varies from 30% to 35% [10,22], over 42% [16], 50% [3,4], and 56% [23] to 75% [24]. Searching for information related to one’s children’s health can be an expression of increased informational needs in the face of the new responsibilities of parenting [25].

Data about parents’ web searches for information related to their children’s health mirror data about Web searches for health information in general. Studies have reported that the majority of people (>60%) searching online for health information used search engines to find the websites [10,18] and that 75% of parents looking for information about their child’s condition used Google [26]. Concerns about the quality of Web-based health information in general are mirrored by concerns about websites regarding children’s health issues [14,27-30] including vaccination [31,32], other procedures [33], specific conditions such as prematurity [34], illnesses such as eczema [35], and more general health advice, such as recommendations for children’s toothbrushing [36].

The reported benefits of searching online for information related to one’s children’s health include being better informed about health and health care provision [37], the availability of information outside doctors’ hours [25], decreased use of services for minor, nonurgent issues [38], and improved decision making in urgent cases [38]. Sometimes parents are even able to arrive at the correct diagnosis when the attending pediatrician has failed to do so [39]. The vast majority (89%‑90%) of parents who have used the internet to find information related to their children’s health found it useful [26,40,41] and many (67%) said it had influenced the medical decisions made on behalf of their child [40]. Interestingly, only a minority of parents chose to discuss the results of their Web search with their child’s doctor [26].

However, despite the clear benefits of searching the internet for health information, 18% [26] to 30% [42] of parents reported finding the websites confusing and distressing or felt more anxious after their search. In other words, in a small number of families, parents’ search for Web-based health information can have detrimental effects. Apart from reports of cases where attempts to implement advice obtained online have gone wrong [27], parents’ searches can negatively affect communication with health professionals [22,42-45] and lead to distress and anxiety for the parents [22,26,42]. In the case of searches for information regarding one’s own health, a similar pattern of behavior, termed “cyberchondria,” has been described [46-51]. Cyberchondria is characterized by Web searches for information about perceived symptoms and possible diagnoses. The searches are undertaken with the aim of assuaging worries and relieving anxiety, yet they increase them and potentially lead to further searches in a quest for reassurance. The increased worry may be a consequence of the search itself, the material encountered online, or the accompanying attentional and cognitive processes. Cyberchondria is a multidimensional construct with five main facets: compulsion, ie, Web searches interrupt activities of daily living; distress, ie, search behavior has negative emotional consequences; excessiveness, ie, repeated searching and, sometimes, escalating search behavior; need for reassurance, ie, desire to consult a health care professional after the Web search; mistrust of medical professionals, ie, putting more faith in the information one has researched by oneself than in trusting expert opinion. Cyberchondria is related to health anxiety [52,53]. The vicious circle that occurs when searching for health information for oneself [46] may also apply when searching for information related to one’s children’s health. Some authors call this pattern of behavior “cyberchondria by proxy” [22]. We do not recommend usage of this term as it suggests a resemblance to the Munchausen syndrome by proxy and thus appears unnecessarily stigmatizing, especially given that hardly anything is known about the consequences of such parental searches. As a neutral phrase, we suggest using “excessive search for Web-based information related to one’s children’s health.”

The aim of the study was to improve knowledge regarding parents’ excessive Web-based searching for information related to one’s children’s health by developing an instrument to capture such search behavior. We also aimed at investigating whether such behavior is associated with personal health anxiety and with searching for Web-based health information related to one’s own health, the perceived health status of one’s child, and other relevant health information, such as a chronic illness of the child or of the parents.

The study was conducted according to the Declaration of Helsinki and approved by the Internal Review Board of Göttingen University (number 2017-162). Data were collected anonymously, and participants were not able to access the survey until they had viewed information about the study and provided consent.

The first step was the generation of an item pool through brainstorming by the researchers and advanced students familiar with the construct of cyberchondria. We reviewed the existing research literature as referenced in the Introduction section. As no instruments exist to assess excessive searching for Web-based information related to one’s children’s health, we especially considered research in the field of adult cyberchondria [47-52] and previously established measurement instruments for cyberchondria such as the Cyberchondria Severity Scale (CSS) [48] as the basis for item development. Items of the CSS were assessed with regard to their suitability for parents’ Web searches, and select item content was adapted for the item pool. In addition, informal interviews were conducted with parents of children aged 0 to 10 years, clinical psychologists specialized in treating children and adolescents, nursery school teachers, and pediatricians. This resulted in a pool of 84 items that were screened to remove items regarded as too linguistically complicated (eg, double negations, uncommon words) or redundant, leaving a final pool of 63 items. The 63 items were presented, in a randomized order, to an online sample of parents of children between the ages of 0 and 10 years. The parents were recruited online through parenting forums and advertisements on notice boards. A total of 486 parents, almost all being mothers (471/486, 96.9%), completed the questionnaire. Their mean age was 31.9 years (SD 5.3). After an analysis of the distribution of missing answers, item difficulty, and participant comments, exploratory factor analyses (EFAs) and reliability analyses were conducted. Items were selected in an iterative process with items excluded if they showed unsatisfactory item characteristics.

The final instrument, Children’s Health Internet Research Parental Inventory (CHIRPI), consisted of 21 items organized into three 7-item subscales. The Distress subscale captures the aversive emotional consequences of parents’ Web searches for information related to their children’s health, such as anxiety or irritation, captures physiological arousal, such as difficulty falling asleep or finding it hard to relax, and captures the interference of these consequences with everyday activities. The Symptom Focus subscale captures parents’ search behavior that is prompted by a perceived physical change in their children and is aimed at finding the causes for the symptoms or a cure for a presumed disease. The Implementing Advice subscale captures parents’ tendency to implement Web-based advice and encompasses various behaviors, such as asking the child’s doctor to prescribe specific medication or carry out particular diagnostic procedures.

Participants who agreed to take part in a retest received an email invitation with a link to a short version of the Web-based survey 3 weeks after they had first completed it. They repeated basic demographic information (described in the Demographic Information About Parents and Children section), health-related information (described in the Health-Related Information section), and CHIRPI.

We conducted a standard item analysis for CHIRPI, first inspecting the number of missing data points for each item and calculating item means, standard deviations, and difficulties. Interitem correlations and item-total correlations were also computed.

We investigated the factorial structure of CHIRPI using an EFA. The suitability of the data for a factor analysis was ascertained using the Kaiser-Meyer-Olkin (KMO) criterion and Bartlett’s test. The number of factors to be extracted was determined using a Velicer minimum average partial (MAP) test [61]. Factors were extracted using a maximum likelihood estimation, and the extracted factors were rotated obliquely (Promax with a Kaiser normalization).

Reliability analyses were carried out by computing the internal consistency (standardized alpha) of the whole scale and the subscales and examining whether the scale and subscales would have benefitted from the removal of individual items. The retest reliabilities of the whole scale and the subscales were calculated as Pearson correlations between the first and second measurements. Parents who completed the retest too late (more than 5 weeks after first completing CHIRPI) were excluded from the retest. We also used independent sample t tests (2-tailed) to determine whether the subsample that had agreed to take part in the retest differed from the subsample that chose not to participate with regard to age, number of children, time spent searching for Web-based information related to their children’s health, and scores in CHIRPI, CSS-15, mSHAI, and CVS.

The correlations were calculated between the CHIRPI scale and its subscales and the demographic information, global (VAS) ratings regarding worry about the child’s health, perceived parental medical competence, perceived general health of the child, and previous experience with the child’s doctors. In addition, we calculated the correlations between CHIRPI scores and scores in CVS, mSHAI, and CSS-15.

We used a 2×2 analysis of variance (ANOVA) with the factors child’s health status (healthy/chronically ill) and the perceived health vulnerability of the child (CVS−: CVS below cutoff/CVS+: CVS above cutoff), and the CHIRPI total score as a dependent variable to investigate the influence of the child’s objective health status and the parent’s perceptions of the child’s vulnerability on the CHIRPI score. These analysis was repeated for the subscales. As a measure of effect size, we reported eta-squared [62]. The data were analyzed using Statistica (TIBCO Software Inc, 2018, version 13, Palo Alto, CA), except for the EFA, which was conducted using SPSS (IBM, version 25, Armonk, New York).

The Web-based survey was widely publicized on German webpages and social media related to parenthood and children. Participation was open to all interested parents, and the resulting sample constituted a convenience sample. Participants were included in the analysis if they reported German as their native language and had at least one child under 10 years of age. A total of 515 participants gave informed consent, and 491 of the participants provided the basic data required to assess eligibility. The inclusion criteria were fulfilled by 480 respondents (nine reported another native language and in two cases the youngest child was older than 10 years). A total of 92 of the 480 respondents (92/480, 19.2%) stopped responding before they had completed CHIRPI. Furthermore, two participants provided implausible data and had to be excluded.

Data from a final set of 384 participants (372/384, 96.9% mothers and 12/384, 3.1% fathers) were analyzed. The mean age of the sample was 32.7 years (SD 5.8) and they had an average of 1.8 children (186 participants had 1 child, 137 participants had 2 children, 40 participants had 3 children, 13 participants had 4 children and 8 participants had five or more children). The mean age of the participants’ youngest child was 3.4 years (SD 3.0). Approximately one in 10 (37/384, 9.6%) of the participating parents reported being single parents and a quarter (94/384, 25.4%) worked in a health-related profession. In total, 11.7% (45/384) reported that at least one of their children had a chronic illness, 22.9% (88/384) said they personally had a chronic illness, and 15.9% (61/384) said their partner suffered from a chronic illness; 37.8% (145/384) reported that at least one member of their immediate family had some form of a chronic illness.

There were very few missing responses in CHIRPI. The frequency of missing responses ranged from 0 (most items) to 3 (item 20; see Table 1). Item 16 was the most difficult, ie, least endorsed (p_i=0.07) and item 2, the easiest, ie, most endorsed (p_i=0.65). The mean item difficulty was p_i=0.29, and the interitem correlation for the whole scale was r=0.29. Item-total correlations ranged from r_itc=0.37 (item 20) to r_itc=0.65 (item 10), with a mean of r_itc=0.50.

The KMO criterion value of 0.88 and the Bartlett test chi-square value of 3421.3 (P<.001) indicated that the data were suitable for an EFA. The Velicer MAP test suggested the extraction of three factors. The maximum likelihood factor estimation converged after six iterations and three factors were extracted, which together explained 45.16% of the variance in the CHIRPI scores. The factor loadings ranged from 0.42 to 0.86, and no item showed double loadings of >0.30 with another factor (Table 2). Factor 1 aggregated items that were indicative of distress as a consequence of searching and was called “Distress.” Factor 2 was related to a symptom-focused search strategy and was named “Symptom Focus.” Factor 3 encompassed various behaviors aimed at implementing advice found online and was called “Implementing Advice.” As the factors showed good internal consistency, we treated them as subscales. The subscales were moderately correlated, Distress×Symptom Focus: r=0.36 (P<.001); Distress×Implementing Advice: r=0.44 (P<.001); Symptom Focus×Implementing Advice: r=0.50 (P<.001).

CHIRPI is the first instrument to assess parents’ excessive Web-based searching for information related to their children’s health. The development of the inventory was guided by previous research on excessive Web-based research for health information [48-52], participative interviews with parents and children’s health professionals, and psychometric standards. The inventory was validated in a good-sized online sample consisting mostly of mothers. Acceptance of the scale was high, and its psychometric characteristics are good to excellent.

Regarding the quality of the individual items, the general feedback in the validation and pilot tests indicated that the items were understandable and well accepted. The extremely low number of missing responses to individual items also suggested that participants did not find CHIRPI hard to complete. The statistical investigation showed that the item difficulties and item-total correlations were good to very good. In the context of the measurement of attitudes and behaviors, item difficulty is a measure of how frequently the statement made in an item is endorsed. Endorsement rates below 0.20 indicate comparatively low agreement, whereas endorsements above 0.80 point to widespread agreement. The endorsement rates for CHIRPI items ranged from 0.07 (“I ask my child’s doctor to prescribe particular drugs that I have read about online.”) to 0.65 (“When I search online for health information relevant to my child, I am particularly interested in whether the problems or symptoms need treatment.”), with a mean of 0.29. For maximum diagnostic value, item difficulties should range from 0.20 to 0.80. However, considering that we are aiming to assess a behavioral excess, ie, a behavior that is relatively uncommon, the scale should differentiate better in the area of low endorsement. Therefore, the range of difficulties in CHIRPI seems to be appropriate. Item-total correlations capture the relationship between an individual item and the scale as a whole and those of the CHIRPI ranged from 0.37 to 0.65, with a mean of 0.50. These values are medium to high; in general, item-total correlations of 0.30 to 0.50 are regarded as medium and correlations of 0.50 and above, as high.

The factor analysis revealed a clear factor structure without any double loadings, consisting of 3 moderately correlated factors, which together explained 45% of the variance. The first factor, Distress, captures how distressed a parent feels as a result of Web searches for health information. It captures cognitive (worry; item 21), emotional (feeling anxious, panicky, angry, or irritated; items 7, 9, and 12), and behavioral (finding it hard to relax or sleep or interference of the search with daily activities; items 4, 15, and 18) aspects of distress. The second factor, Symptom Focus, relates to the characteristics of the search behavior. It captures search behavior focused on changes the parent has noticed in his or her child’s body and on attempts to determine the assumed causes of the symptoms or the presumed underlying illness. The third subscale, Implementing Advice, aggregates behaviors such as following general health advice for one’s child that one has read about online. Examples for such use of internet advice may be the implementation of diets or exercise regimes, administering nonprescription medicine, asking the doctor for prescription drugs or diagnostic procedures, and engaging teachers or the nursery staff’s assistance with observing symptoms or implementing health regimens.

It is important to note both similarities and differences of CHIRPI to the best-established instrument for assessing cyberchondria in adults searching for Web-based health information for themselves [48], especially as CSS-related research informed the generation of CHIRPI. Similar to the CSS, CHIRPI encompasses a Distress Scale, which consists of adapted item content of the respective CSS for parents’ Web searches. It seems that this factor, reflecting the distress after searches, plays an important role in both constructs. In contrast to the CSS, however, CHIRPI hardly contains references to the compulsive nature of the Web searches or the mistrust toward medical professionals. Although items pertaining to this content were part of the original item pool for CHIRPI, they did not perform well during the questionnaire development phase and were thus eliminated by the iterative EFA process—perhaps because these facets are of less importance to parents’ health-related Web searches. Excessiveness of the searches and reassurance by these searches were subsumed under the CHIRPI subscales, Symptom Focus, and Implementing Advice.

We investigated two aspects of reliability: internal consistency and the 4-week retest reliability. The internal consistency of the total scale was high, with standardized alpha=.89, especially given its modest length (21 items). The subscales also showed good (Distress alpha=.89; Symptom Focus alpha=.87) or satisfactory (Implementing Advice alpha=.74) internal consistency. The comparatively low internal consistency of the Implementing Advice subscale might reflect the greater heterogeneity of behaviors described, which is indicated by the lower interitem correlations. Whereas the items of the Distress subscale capture aspects of distress that often co-occur (eg, increased worry, finding it harder to relax, problems with sleep), the items that form the Implementing Advice subscale describe behaviors that—albeit all indicative of the tendency to act on Web-based advice—need not co-occur. Participants who ask nursery staff or teachers for help with observing symptoms may not also ask their child’s doctor to prescribe medications they learned about online.

Test-retest reliability is a measure of the degree of correspondence between questionnaire responses given at different points in time and is influenced by the quality of the questionnaire on the one hand and the stability of the measured construct on the other. At present, little is known about the temporal stability of health-related Web-based search behavior regarding one’s children, in general, or excessive searching for such information, in particular. The 4-week retest reliability of r_tt=0.78 of CHIRPI suggests the phenomenon is stable across a medium time-frame of weeks and indicates that the instrument has good test-retest reliability. The test-retest reliability was highest for the Distress subscale (r_tt=0.84), which suggests that the tendency to experience worry and distress as a consequence of the searching may be an enduring trait. However, although the good test-retest reliability suggests the construct is reasonably stable and that the instrument used to measure it has adequate psychometric characteristics, further research in both these areas is warranted. In particular, it would be instructive to investigate the sensitivity of CHIRPI to change—eg, after an intervention aimed at reducing excessive Web searches by parents. In studies such as ours, where the subsample for the retest is self-selecting, the retest sample may be biased (eg, toward respondents particularly affected by the phenomenon under investigation), so we compared the data of the participants that formed our retest subsample against those who did not provide data for the retest. The samples did not differ with regard to the CHIRPI scores or scores on any of the other questionnaires, but parents with more children were less likely to participate again (small effect), possibly reflecting greater competition for their time.

We investigated relationships between CHIRPI and related constructs, such as parental cyberchondria and worry about one’s own health and the health of one’s children to provide insights into the instrument’s validity. In addition, we investigated, how the inventory performed in particular groups of parents, namely parents who reported that one of their children had a chronic illness and parents who perceived their child’s health to be vulnerable (based on the CVS score).

When interpreting the results, the following limitations must be borne in mind. First, our sample consisted predominantly of mothers; however, reflecting offline behaviors [64,65], the majority of those searching online for health information are women [2,10,18,66-68]. Second, our data are cross-sectional, so we cannot draw inferences about causality. Third, all our data are self-reported and may have been affected by a self-selection bias.

Most parents search on the internet for information related to their children’s health and use it to their advantage. However, for a minority of parents, searching may escalate and have detrimental consequences similar to self-directed cyberchondria. Health professionals working with children worry that parents’ Web searches for health information may affect their own role and their relationships with parents and children [69]. CHIRPI allows us to investigate such issues as it captures various forms of parents’ excessive search behavior (eg, symptom-focused search) and their consequences, such as distress after the search or the implementation of newly acquired advice. CHIRPI has the potential to help identify parents who are at risk of child-directed cyberchondriac behavior. More research is needed to determine the potentially negative consequences of excessive searching for Web-based information related to one’s children’s health, eg, increased use of health services, children being subjected to unnecessary diagnostic procedures, and the transmission of unsubstantiated health concerns. The Distress subscale of CHIRPI, in particular, seems to be a useful marker of dysfunctional search influences, such as parents’ anxiety about their own health or self-directed cyberchondria. It may indicate general patterns of factors related to anxious health-related cognitions, emotions, and behaviors, which the parents also apply to the health of their children. Hopefully, the availability of CHIRPI will improve our understanding of the consequences of excessive Web-based searching for health information.

Primarily, CHIRPI can be used to gain greater knowledge of health-related Web searches by parents and the possible consequences of such search behavior. The authors do not wish to suggest that searching for such information is problematic in itself. However, if it is coupled with health anxiety and cyberchondria regarding the parents themselves, it may become excessive and a source of distress for the parents. We would therefore suggest using this instrument to assess whether a person being treated for health anxiety also searches for their children and, if appropriate, discuss this behavior in the course of the treatment. Future research should address whether parents’ excessive Web searches for health information regarding their children also leads to problematic consequences for the children themselves.