1 Introduction

In recent years, the global percentage of Internet users has grown exponentially and smartphone has become the main device in its access (We Are Social and Hootsuite, 2022). The expansion of Internet access has changed the way people live, work, communicate, and learn and has become an essential environment in their development (Pinho et al., 2021). In the education sector, incorporation of the use of Internet services has led to multiple improvements in the teaching–learning process (Wu et al., 2010), and specifically at universities, has allowed eliminating geographical barriers and increasing flexibility (Santhanam et al., 2008; Yakubu et al., 2020).

However, university students do not only use the Internet for educational and academic purposes, but also to look for information, random navigation, entertainment, communication, gaming, social networks, and online shopping and, to a lesser extent, gambling and obtaining sexual information (Adorjan et al., 2021; Anand et al., 2018; Balhara et al., 2019; Maqableh et al., 2021; Servidio, 2014; Zenebe et al., 2021). These coincide with the purposes of smartphone use by this population (Coban & Gundogmus, 2019; Matar Boumosleh & Jaalouk, 2017).

The widespread availability of the Internet through smartphones and other devices is associated with multiple benefits, such as access to information and a space for social communication and entertainment. (e.g., Maia et al., 2020; Manago et al., 2012). However, Internet penetration in everyday life is a serious problem for an increasing number of people, rising to the level of problematic Internet use (PIU) or problematic smartphone use (PSU). These problematic behaviours are associated with negative consequences such as poor academic performance (Anderson et al., 2017; Grant et al., 2019), psychological distress (Busch & McCarthy, 2021; Chen et al., 2020; Odacı & Çikrikci, 2017; Radeef & Faisal, 2018; Weinstein et al., 2015), and disturbed sleep and daytime sleepiness (Ferreira et al., 2017; Yang et al., 2020a), to name a few.

1.1 Problematic Internet and smartphone use

Although there was already concern about addictive use of the internet by the end of the last century (Griffiths, 1995; Young, 1998a), today, there is currently greater recognition of technological addictions in mental health by both the American Psychiatric Association (APA, 2013) and the World Health Organization (WHO, 2018), and excessive use of digital technologies has been recognised as a public health issue (WHO, 2015).

Today, several terms are often used to describe the phenomenon, such as addiction (Young, 1998a) or internet dependency (Dowling & Quirk, 2009). Among these, the term “problematic Internet use” (hereinafter, PIU) stands out, and it is defined as a pattern of maladaptive Internet use characterised by loss of control, the appearance of negative consequences, and obsessive thoughts when the Internet cannot be accessed (D'Hondt et al., 2015). This is an umbrella term (Fineberg et al., 2018) and accommodates the broad spectrum of non-adaptive behaviours online, which go beyond behavioural addiction (Billieux et al., 2015b; Starcevic, 2013). However, the terms "Internet addiction" and PIU are used inconsistently in the literature (Sánchez-Fernández et al., 2022).

Smartphone, because it is portable and gives easy access to internet, has the potential to create high dependency and is a powerful risk factor for problematic and addictive behaviours (Aljomaa et al., 2016; Carbonell et al., 2018). As a result, problematic smartphone use (PSU) is now being discussed more and more, such as excessive smartphone use that interferes with various areas of a person's life (Billieux et al., 2015a).

The current debate is whether PSU can be considered a sub-category of PIU or whether it is an independent phenomenon (Cheever et al., 2018). Recent studies have found that PIU and PSU overlap in some, but not all key features (Lee et al., 2020; Tateno et al., 2019), while others have established that these problematic behaviours all overlap (Kittinger et al., 2012; Montag et al., 2015).

Recent epidemiological studies have found a large variability in the prevalence rates of PIU and PSU in the general population (López-Fernández & Kuss, 2020; Sohn et al., 2019). In the case of university students, variability has also been found in the prevalence rates of PIU (4—51%), which may be explained by the lack of diagnostic criteria and cultural differences between samples (Sánchez-Fernández et al., 2022). However, despite this variability, these problems increase over time (López-Fernández & Kuss, 2020; Kuss et al., 2021; Pan et al., 2020; Shao et al., 2018) and university students tend to be at higher risk of PIU (Anderson et al., 2017; Ferrante & Venuleo, 2021; Kuss et al., 2014) and PSU (Roig-Vila et al., 2020).

1.2 Specific problematic Internet uses

PIU/PSU is a broad term that may include a variety of problematic behaviours. In fact, individuals who use the Internet/smartphone excessively do not become addicted to the Internet/smartphone environment but to the behaviours they engage in when they are online (Király & Demetrovics, 2021; Meerkerk et al., 2009). That is why some authors are sceptical about the viability of PIU/PSU as a construct, and favour the examination of specific activities such as playing games or sexual activity (e.g., Starcevic & Aboujaoude, 2017). At the beginning of the century, Davis (2001) made a distinction between two different forms of pathological use of the Internet: general and specific. The general one includes a broader set of behaviours while the specific one refers to engagement with specific Internet functions or applications. Years later, Billieux (2012) argued for the existence of a spectrum of cyber addictions that would include problematic behaviours related to the smartphone, in general, and specific online activities such as video games and online gambling, pornography, and social networks. More recently, some authors have conceptualised problem behaviours mediated by the Internet and smartphones as being within a spectrum of related conditions associated with both shared and unique characteristics (Baggio et al., 2018). In this study, "specific problematic Internet uses" will refer to those problematic online behaviours that can be carried out via smartphone or any other device.

Previous reviews (Kuss et al., 2021; Lopez-Fernandez & Kuss, 2020) have established four main themes in terms of specific PIU: problematic social media use (PSMU), Internet gaming disorder (IGD), problematic Internet pornography use (PIPU), and problematic Internet gambling. These studies show great variability in terms of terminology (addiction coexisting with problematic use, disorder, or dependence, among others), as well as in measurement instruments and prevalence rates. So far, no systematised data focusing on the university student population have been found.

1.3 Conceptualisation of problematic Internet use behaviours

A number of diverse etiological models have been proposed in the conceptualisation of these problem behaviours (Ferrante & Venuleo, 2021). Davis (2001), in his cognitive behavioural model of pathological Internet use, proposes that psychopathology (distal cause) would give rise to the PIU, generalized or specific, through maladaptive cognitions (proximal cause such as low self-efficacy or negative self-evaluation). The behavioural symptoms of PIU are reinforcements of the maladaptive cognitions that result in a vicious circle and maintain pathological behaviour.

On the other hand, the person-affect-cognition-execution (I-PACE) model (Brand et al., 2016, 2019) argues that specific problematic uses of the Internet are the consequence of interactions between predisposing (personality-related characteristics, social cognitions, biopsychological factors and motivation to use), moderating (coping styles and Internet-related cognitive biases) and mediating (affective and cognitive responses to situational triggers) factors in combination with reduced executive functioning. These associations would be maintained by Pavlovian and instrumental conditioning processes within an addiction process. The authors also assume that the medium (internet, smartphone) is secondary in the origin of these problem behaviours and that, among the psychological and neurobiological mechanisms, some are common and others specific to each addictive behaviour (such as specific personality profiles) (Brand et al., 2016, 2019).

1.4 Risk and protective factors for problematic internet usage behaviours

Research on shared risk and protective factors unique to the spectrum of online problem behaviours is essential for advancing their conceptualisation and prevention, which in turn will have clear implications for the overall health and well-being of university students (Tugtekin et al., 2020). Problem behaviours mediated by the Internet and smartphones are associated with both shared and unique risk factors (Baggio et al., 2018; Billieux, 2012).

Previous reviews have examined risk factors for PIU, finding that being younger, being male, a higher family socio-economic status, duration of use, social networking and gaming, neuroticism, impulsivity, loneliness, depression, anxiety and general psychopathology increase the risk of generalized PIU (Aznar Díaz et al., 2020; Kuss et al, 2014, 2021); depression and aggression were the main risk factors for online gaming addiction, Internet gambling problems were associated with lower emotional intelligence and psychological distress, and problematic online pornography use was most frequently related to relationship problems, disruptive worry and behavioural dysregulation (Kuss et al., 2021). On the other hand, with regard to PSU, the review by Wacks and Weinstein (2021) concludes that it is associated with psychiatric, cognitive, emotional, medical and brain alterations. For their part, Busch and McCarthy's (2021) review of the predisposing factors to PSU found a great variety of backgrounds divided into four categories: Control (e.g. self-control or tolerance of uncertainty), Emotional health (e.g. anxiety and depression), Physical health (e.g. Individual's health status), Preconditions (e.g. family characteristics), Professional performance (e.g. academic performance), Social performance (e.g. personality) and Technology features (e.g. type of mobile phone use).

However, no consistent findings have been found in research on predictors of generalized and specific PIU and PSU in the university student population.

Prior to this study, the authors reviewed studies on risk and protective factors for generalized PIU in university students. Ten predictive factors for PIU have been identified and divided into three categories (patterns of use, psychological variables, and lifestyles). Among these, nine were risk factors (time spent online, video games, depression, negative affect, life stress, maladaptive cognitions, impulsivity, poor sleep quality, substance use (alcohol and drugs), and one was a protective factor (conscientiousness). However, all studies that focus on other technologies-related problem behaviours, such as PSU or specific PIU, were excluded from this review.

1.5 The purpose of the study

Consequently, the aim of this systematic review is to examine the studies of predictors of PSU and specific PIU (online gaming, social networking, online gambling, and online pornography) in university students that have been published since the inclusion of IGD in the DSM-5.

The research questions are: 1. What terminology is used to refer to PSU and specific PIU?, 2. What are the assessment tools used in the PSU and specific PIU evaluation?, 3. What is the prevalence of PSU and specific PIU?, 4. What are the risk and protective factors associated with PSU and specific PIU? From these questions, the objectives are as follows: 1. To become familiar with the terminology used, 2. To review the assessment tools, 3. To analyse prevalence and 4. To study the risk and protective factors associated with PSU and specific PIU.

2 Methodology

Systematic review methodology was used (Page et al., 2021). We included scientific research articles published between 2013, the year when "Internet gaming disorder" (IGD) in DSM-5 (APA, 2013) was officially recognised, and the expansion of smartphone use (Carbonell et al., 2018), and 2021, both years included, on predictive, risk and protective factors associated with PSU and specific problematic internet use (e.g., social networking) in university students. The Web of Science, Proquest (PsycINFO and Medline) and Scopus databases were used between October and December 2021. The keyword strategy used the terms, clusters, and Boolean operators listed in Table 1 (also translated into Spanish, French, and Italian). The search was done by article title, abstract and keywords.

Table 1 Search strategy
Full size table

2.1 Inclusion and exclusion criteria

The inclusion criteria were: (1) scientific articles; (2) study factors predicting PSU or specific PIU through predictive modelling; (3) university students from more than one area of knowledge; (4) 17 years or older; (5) use of validated instrument; (6) quantitative empirical data; (7) reported effect size; (8) access to full text; and (9) written in English, Spanish, French and Italian.

The exclusion criteria were: (1) studying predictors of PIU or other behavioural addictions that do not involve Internet use (e.g. offline gambling) or substance-related addictions (e.g. alcoholism); (2) lack of relevant data; (3) non-university sample; (4) under 17 years of age; (5) university students from a single field of knowledge; (6) sample collected during the covid-19 lockdown; (7) no predictive statistical model; (8) no use of validated instruments; (9) validation studies of assessment instruments; (10) unreported effect size; and (12) sources other than peer-reviewed journals (e.g. non-peer-reviewed journals, conference abstracts, chapters, books, corrections).

2.2 Selection of articles

The PRISMA protocol guidelines were followed (Page et al., 2021) (See Fig. 1). Our initial sample contained 117 articles. They were included in the present review after screening duplicates and articles that met the inclusion criteria mentioned, but not the exclusion criteria. The included studies were organized into four main subjects, one of them referring to a generalized problematic use of smartphones: Problematic Smartphone Use (PSU) and three on specific problematic uses of the internet: Problematic Social Media Use (PSMU), Internet Gaming Disorder (IGD) and Problematic Internet Pornographic Use (PIPU). Since they analysed more than one variable, some studies have been repeated in two subject groups.

Fig. 1
figure 1

PRISMA Diagram of study selection processes

Full size image

At a second stage, articles whose predictive factors were supported by at least 3 studies were chosen, in order to address Objective 4. As a result of this second analysis, 83 articles were selected that studied 10 PSU factors, 5 PSMU factors, and 1 IGD factor.

2.3 Data extraction

The characteristics of the 117 studies selected are shown in Table 2. In terms of effect size, betas (ß), odds ratio (OR) and coefficient of determination (R2) were included. For betas, the cut-off points are used: Very small > 0 to < 0.1, small ≥ 0.1 to < 0.3, medium ≥ 0.3 to < 0.5 and large ≥ 0.5 (Cohen, 2013, Ferguson, 2016). For odds ratios (OR): Very small > 0 to < 1.5, small ≥ 1.5 to < 2, medium ≥ 2 to < 3, and large ≥ 3 (Sullivan & Feinn, 2012). With respect to the coefficient of determination (R2): Very small < 0.02, small ≥ 0.02, medium ≥ 0.13 and large ≥ 0.26 (Dominguez-Lara, 2017).

Table 2 Descriptive characteristics of selected articles in alphabetical order (N = 117)
Full size table

2.4 Quality assessment

The "quality assessment tool for observational cohort and cross-sectional studies" was used (National Heart, Lung, and Blood Institute, 2014). 50.4% of articles (n = 59) were assessed by the first and second authors. After obtaining an adequate interjudge agreement (98.9%), the first author completed the quality assessment.

Articles are classified as "good", "fair" or "poor" quality based on an overall judgment based on 14 criteria (Table 3). Four are not applicable to cross-sectional studies (6, 7, 10 and 13). All meet the 4th and 9th criteria (university student, over 17 years old and measure problematic behaviour by validated questionnaire). All meet the 12th criterion. They were filled in anonymously. Compliance with the 2nd criterion (demographics, location and time period when they were obtained) is considered to be of good quality. For the 8th criterion, a dichotomized measurement of the independent variable is considered to be of good quality if the cut-off point used is validated. For the 14th criterion, a model controlling the variables of sex (if not stratified), age (if not in the range ± 1 year), and a third confounding factor (for example, socioeconomic status or academic performance) is considered sufficient. 38.5% of the articles have not reported participation rates, preventing the 3rd criterion from being assessed.

Table 3 Checklist criteria, from the U.S. Department of Health & Human Services
Full size table

2.5 Data analysis

In Objectives 1, 2 and 3, a descriptive analysis of the 117 articles included was carried out (See Table 2). Objective 4 analysed all 83 studies, examining 17 predictive factors included in the review.

3 Results

The database search identified 117 studies that examined risk and protective factors for problem behaviours online (see Table 2).

In the quality assessment, 56 articles (47.9%) were of good quality, 30 (52.1%) were medium, and none displayed poor quality. The most frequent limitations were lack of complete definition of the population (2nd criterion), non-contemplation of different levels of exposure of potential predictive factors (8th criterion), and no introduction of confounding variables in predictive models (14th criterion).

The articles included have been divided into four groups based on specific online behaviour: PSU (n = 67), PSMU (n = 39), IGD (n = 9), y PIPU (n = 2). These five constructs are treated separately because research suggests that their prevalence rates and risk factors appear to be different from each other (Billieux, 2012; Kuss et al., 2021).

3.1 Problematic smartphone use

A total of 67 studies have analysed the predictive factors for PSU in college students.

3.1.1 Description of studies

The design was longitudinal in 5 studies (7.5%) (Cui et al., 2021; Elhai et al., 2018b; Rozgonjuk et al., 2019; Yuan et al., 2021; Zhang et al., 2021) and in 62 it was cross-sectional (92.5%). 75% were from Asia (n = 50) and 16.4% from North America (USA) (n = 11), and a smaller number were from Europe (n = 4) and South America (n = 2, 3%). 29.8% of the studies were published between 2013—2017, and 70.2% from 2018. The samples ranged from 191 (Roberts and Pirog III, 2013) to 2367 students (Alosaimi et al., 2016), with 82.1% below 1000 and a mean of 617.3 (SD = 392.8).

The terms ‘smart/mobile/cell phone addiction’ (n = 31, 46.3%), ‘problematic smart/mobile phone use’ (n = 30, 44.8%), ‘smartphone use disorder’ (n = 2, 3%), ‘mobile phone dependence’ (n = 2, 3%), ‘excessive smartphone use’ (n = 1 1.5, 1.5%) and ‘smartphone dependency’ (n = 1%) were used.

17 assessment instruments were identified (see Table 4). High scores indicated a higher degree of PSU. The most widely used was the Smartphone Addiction Scale (SAS) (Kwon et al., 2013) (n = 24, 35.8%). Different versions of this instrument have been used. 14 used the short version (range: 10—60) with means between 20.96 (SD = 7.56) in Turkey (Gökçearslan et al., 2016) and 41.35 (SD = 35.95) in Italy (De Pasquale et al., 2019). Eight used the standard version (range: 33—198) with means between 68.46 (SD = 24.95) in South Korea (Choi et al., 2015) and 98.00 (SD = 26.73) in Brazil (Laurence et al., 2020), and 2 used the undergraduate version (Su et al., 2014) (range: 1—5) with means between 2.85 (SD = 0.55) (Lin et al., 2021) and 2.88 (SD = 0.566) (He et al., 2020) both in China. Ten studies used the ‘Mobile Phone Addiction Index Scale’ (Leung, 2008) (range = 1—5) with averages between 1.99 (SD = 0.58) (Zhang et al., 2020a) and 2.78 (SD = 0.72) (Yang et al., 2020b), both in China.

Table 4 Assessment tools
Full size table

Eleven studies (16.4%) reported the prevalence. Thus, the study by Yang et al. (2021a) using the 'Mobile phone addiction tendency scale (MPATS)', with a total score of 16 to 80 (a higher score indicating a deeper degree of addiction), and dividing the total sample into three groups (from 16 to 31 is classified as "no mobile phone addiction", from 32 to 56 is classified as "possible mobile phone addiction", and those equal to or higher than 57 are classified as "mobile phone addiction"), found that 78.29% were classified as possible addiction and 8.06% as addiction. In contrast, the study by Jiang and Shi (2016), using the ‘Problematic Mobile Phone Use Scale (PMPUS)’ and making a dichotomous division, finds that 8.99% display problematic use. Meanwhile, the study by Gündoğmuş et al. (2021), using 'The Smartphone Addiction Scale-Short Version (SAS-SV)', with a score between 10 and 60, and making a dichotomous division (yes/no) finds 48.6% of participants fall within the "addiction" group (SAS-SV ≥ 31 for boys and SAS-SV ≥ 33 for girls). The study by Long et al. (2016) used the Problematic Cellular Phone Use Questionnaire (PCPUQ), a questionnaire composed of 12 items, the first seven of which asked whether in the previous year the participants had symptoms of problematic CPU, while the last five determined the subjective functional impairment of the participants in the previous year caused by CPU, so that participants who had positive responses to four or more of the first seven questions and those who had positive responses to any of the last five questions were classified as having problematic CPU, and found a prevalence of 21.3%.

3.1.2 Predictive factors

From the 67 studies of adequate quality, 10 potential predictive factors of PSU were extracted, classified into two categories (Internet use patterns and psychological variables) (see Table 5).

Table 5 Problematic smartphone use predictive factors
Full size table

Internet use patterns

Smartphone use (time, frequency). Out of 10 studies, all reported that smartphone use was a potential risk factor for PSU (Alosaimi et al., 2016; Elhai et al., 2018a; Elhai et al., 2018c; Elhai et al., 2020b; Erdem & Uzun, 2020; Gökçearslan et al., 2016; Hong et al., 2021; Lin et al., 2021; Wolniewicz et al., 2020). Seven showed a small effect size (70%), two a moderate size (20%) and one a large effect size (10%). The study by Matar Boumosleh and Jaalouk (2017) found a significant effect of excessive mobile phone use (> 5 h per day) but not of daily usage time.

Social media use. Out of 6 studies, all reported that social media use was a potential risk factor for PSU (Abbasi et al., 2021; Coban & Gundogmus, 2019; Koç & Turan, 2021; Lin & Chiang, 2017; Rozgonjuk et al., 2018; Salehan & Negahban, 2013). One showed a small effect size (16.7%), one a moderate size (16.7%), three a large effect size (50%) and one showed a mixed effect (16.7%).

Non social use (process use). Out of 3 studies, all reported that non-social use (process use) was a potential risk factor for PSU (Elhai et al., 2020a; Rozgonjuk & Elhai, 2021; Rozgonjuk et al., 2019). Two showed a small effect size (66.6%) and one moderate (33.3%).

Psychological variables

Negative affectivity (depression, anxiety, stress, boredom proneness, rumination, suicidal ideation). Out of 25 studies, 24 reported that negative affectivity was a potential risk factor for PSU (Alavi et al., 2020; Canale et al., 2021; Choi et al., 2015; Cui et al., 2021; Elhai et al., 2018a; Elhai et al., 2020c; Enez Darcin et al., 2016; Forster et al., 2021; He et al., 2020; Hou et al., 2021; Khoury et al., 2019; Kim & Koh, 2018; Kim et al., 2017; Kim et al., 2019; Kuang-Tsan & Fu-Yuan, 2017; Kuru & Celenk, 2021; Liu et al., 2020; Liu et al., 2021; Long et al., 2016; Matar Boumosleh & Jaalouk, 2017; Xiao et al., 2021; Yang et al., 2020a, 2020b; You et al., 2019; Yuan et al., 2021; Zhang et al., 2020a). Two showed a very small effect size (8.3%), 11 small (45.8%), five moderate (20.8%), one large (4.2%) and five mixed (20.8%).

Specifically, the studies found that depression (Alavi et al., 2020; Choi et al., 2015; Cui et al., 2021; Forster et al., 2021; Kim et al., 2017; Matar Boumosleh & Jaalouk, 2017; Yang et al., 2020a, 2020b; Yuan et al., 2021; Zhang et al., 2020a), anxiety (Alavi et al., 2020; Choi et al., 2015; Hou et al., 2021; Khoury et al., 2019; Kim & Koh, 2018; Kuru & Celenk, 2021; Matar Boumosleh & Jaalouk, 2017), social anxiety (Canale et al., 2021; Enez Darcin et al., 2016; Xiao et al., 2021; You et al., 2019), depression/anxiety and suicidal ideation (Kim et al., 2019), stress (Forster et al., 2021; He et al., 2020; Kim et al., 2019; Kuang-Tsan & Fu-Yuan, 2017; Liu et al., 2020; Long et al., 2016), rumination (Elhai et al., 2020c; Liu et al., 2021) and boredom proneness (Elhai et al., 2018a; Yang et al., 2020a, b; Zhang et al., 2021) were risk factors.

Self-control/regulation. Out of 7 studies, all reported that self-control/self-regulation was a potential protective factor against PSU (Cebi et al., 2019; Gökçearslan et al., 2016; Jiang & Shi, 2016; Pourrazavi et al., 2014; Yang et al., 2019, 2020a, 2020b; Zhang et al., 2021). One showed a very small effect size (14.3%), two small (28.6%), one moderate (14.3%), one large (14.3%) and two mixed (28.6%).

Well-being (self-efficacy, tolerance to distress, self-esteem, vitality, interpersonal adaptation, etc.). Out of 13 studies, 9 reported that well-being was a potential protective factor with respect to PSU (Elhai et al., 2018b; Jiang & Shi, 2016; Kim & Koh, 2018; Koç & Turan, 2021; Lian, 2018; Lian & You, 2017; Sun et al., 2021; Yuchang et al., 2017; Zhang et al., 2020b). One showed a very small effect size (11%), five small (55.6%), two moderate (22.2%) and two mixed (22.2%).

Specifically, the studies found self-efficacy (Jiang & Shi, 2016), social self-efficacy (Sun et al., 2021), distress tolerance (Elhai et al., 2018b), self-esteem (Jiang & Shi, 2016; Kim & Koh, 2018; Koç & Turan, 2021; Yuchang et al., 2017), relationship virtues (Lian, 2018; Lian & You, 2017), interpersonal adaptation (Zhang et al., 2020b) and vitality (Lian & You, 2017),

Neuroticism/emotional instability. Out of 6 studies, three found neuroticism to be a potential risk factor for PSU (Erdem & Uzun, 2020; Liu et al., 2020; Roberts et al., 2015), while two found that emotional stability was a protective factor (De Pasquale et al., 2019; Takao, 2014). These five studies reported a small effect size.

Fear of Missing Out (FOMO). Out of 8 studies, all reported that FOMO was a potential risk factor for PSU (Elhai et al., 2018a; Elhai et al., 2020a, 2020b, 2020c; Handa & Ahuja, 2020; Lin et al., 2021; Wolniewicz et al., 2020; Yuan et al., 2021). Three of them reported a small effect size (37.5%), two a moderate one (25%) and three a large effect size (37.5%). Impulsivity. Out of 4 studies, all reported impulsivity as a potential risk factor for PSU (Canale et al., 2021; Khoury et al., 2019; Roberts & Pirog III, 2013; Roberts et al., 2015). One showed a small effect size (25%), two a small one (50%) and one a mixed effect size (25%).

Other online addictions. Out of 3 studies, all reported that other cyber addictions were potential risk factors for PSU, namely internet addiction (Choi et al., 2015), Facebook addiction (Khoury et al., 2019), and IGD (Yuan et al., 2021). The reported effect sizes were small in two studies (66.6%) and moderate in one (33.3%).

3.2 Problematic social media use

A total of 39 studies have analysed the predictive factors for PSMU in college students.

3.2.1 Description of studies

The design was longitudinal in two studies (5.1%) (Brailovskaia & Margraf, 2017; Brailovskaia et al., 2018) and transverse in 37 (94.9%). 64.1% were from Asia (n = 25) and 23% from Europe (n = 9), and a smaller number were from North America (USA) (n = 4, 10.3%) and Central America (Mexico) (n = 1, 2.6%). 30.8% of the studies were published between 2014—2017, and 69.2% from 2018. The samples ranged from 122 (Brailovskaia et al., 2018) to 1245 students (Hou et al., 2017a), with 94.9% below 1000 and a mean of 433.92 (SD = 240.8).

The following terms were used: ‘social media addiction’ (n = 8, 20.5%), ‘social networking sites addiction’ (n = 4, 10.2%), ‘problematic social media use’ (n = 4, 10.2%), ‘poblematic social networking sites use’ (n = 3, 7.7%) and ‘compulsive social media use’ (n = 1, 2.5%). Others studied the problematic use of certain social networks: ‘Facebook addiction’ (n = 7, 17.9%), ‘problematic Facebook use’ (n = 4, 10.2%), ‘Facebook addiction disorder’ (n = 2, 5.1%), ‘intensive Facebook usage’ (n = 1, 2.5%), ‘WeChat excessive use’ (n = 2, 5.1%), ‘Tuenti addiction’ (n = 1, 2.5%), ‘Instagram addiction’ (n = 3, 7.7%) y ‘Snapchat addiction’ (n = 1, 2.5%).

16 assessment instruments were identified (see Table 4). High scores indicated a higher degree of PSMU. The most widely used was the Bergen Addiction Scale in its different ranges (n = 18, 46.1%). Among them, 13 studies used the Bergen Facebook Addiction Scale (BFAS; Andreassen et al., 2012) with different ranges (1 – 5; 6 – 30; 18 – 90). Out of the 6 studies using the 6—30 range, the means went from 8.98 (SD = 3.64) (Brailovskaia et al., 2018) to 12.88 (4.93) (Balcerowska et al., 2019). The study by Siah et al. (2021) used the Bergen Social Networking Addiction Scale (Andreassen et al., 2012) and four studies used its most up-to-date version, the Bergen Social Media Addiction Scale (BSMAS; Andreassen et al., 2016). Of these latter studies, among those who used a range of 6—30, scores went from 11.96 (SD = 4.99) (Casale et al., 2018) to 16.74 (SD = 4.16) (Chung et al., 2019). On the other hand, 8 studies used specific Generalized PIU instruments such as the Internet Addiction Test (IAT) (Young, 1998a), Generalized Problematic Internet Use Scale-2 (GPIUS-2 2009) (Caplan, 2010), Compulsive Internet Use Scale (CIUS) (Meerkerk, 2007), and the Internet Experience Questionnaire (CERI) (Beranuy et al., 2009).

Four studies (10.2%) reported on the prevalence. So, the study by Punyanunt-Carter et al.  (2018) using the Bergen Facebook Addiction Scale (BFAS; Andreassen et al., 2012), with a total score of 6 to 30, and using a cut-off point of ≥ 18, found that 36.9% had PSMU. Hou et al. (2017a) Using the ‘Excessive WeChat Use scale’ and using two cut-off points (21.4–27.7, > 27.7), reported that 8.2% had ‘excessive use” and 6.6% “serious excessive use”. Kircaburun and Griffiths (2018), using the Internet Addiction Test (IAT) (Young, 1998a) (range: 20–100) and three cut-off points (38–58, 59–73, > 73), found that 26.5% had mild addiction, 6.1% moderate addiction, and 0.9% severe addiction. Jaradat and Atyeh (2017), using the Internet Addiction Test (IAT) (Young, 1998a) (range: 20–100) and using two cut-off points, (50–79, ≥ 80), reported that 62.1% were in the alert group, and 7.9% displayed levels of addiction.

3.2.2 Predictive factors

5 potential predictive factors of PSMU were extracted from the 39 studies of adequate quality, and they were classified into two categories (Internet use patterns and psychological variables) (see Table 6).

Table 6 Problematic social media use predictive factors
Full size table

Internet use patterns

Social media use (time, frequency).Out of 9 studies, all reported that the use of social media was a potential risk factor for PSMU (Chung et al., 2019; Dempsey et al., 2019; Gao et al., 2021; Hong & Chiu, 2016; Hong et al., 2014; Jasso-Medrano & Lopez-Rosales, 2018; Sayeed et al., 2020; Varchetta et al., 2020; Xie & Karan, 2019). Four reported a small effect size (44.4%), three moderate (33.3%) and two a large effect size (22.2%).

Social use. Out of 6 studies, 5 reported that social use of social media was a risk factor for PSMU (Hou et al., 2017a; Kircaburun et al., 2020a; Raza et al., 2020; Sheldon et al., 2021; Süral et al., 2019). Sheldon et al., 2021 found that social activity had a significant effect on Snapchat addiction, with small effect size, but not on Facebook and Instagram addiction. Of the remaining studies that found a significant effect, three reported a small effect size (60%) and one a moderate effect size (20%).

Psychological variables

Negative affect (depression, anxiety, social media communication apprehension, rumination). Out of 10 studies, all reported negative affectivity as a potential risk factor for PSMU (Dempsey et al., 2019; Gao et al., 2021; Hong et al., 2014; Hou et al., 2019; Jasso-Medrano & Lopez-Rosales, 2018; Kircaburun et al., 2020b; Punyanunt-Carter et al., 2018; Sayeed et al., 2020; Xie & Karan, 2019). Nine showed a small effect size (90%) and one a mixed one (10%).

Specifically, the studies found depression (Gao et al., 2021; Hong et al., 2014; Hou et al., 2019; Jasso-Medrano & Lopez-Rosales, 2018; Kircaburun et al., 2020b; Sayeed et al., 2020), anxiety (Gao et al., 2021; Hou et al., 2019; Xie & Karan, 2019), perceived stress (Hou et al., 2017b), Social media Communication Apprehension (Punyanunt-Carter et al., 2018) and rumination (Dempsey et al., 2019) to be risk factors. In addition, Jasso-Medrano and Lopez-Rosales (2018) found a significant and negative effect of suicidal ideation on PSMU.

Well-being (flourishing, life satisfaction, social safeness, relationship satisfaction, trait emotional intelligence, self-confidence, self-esteem, vitality, self-liking, psychological capital, psychological resilience). Out of 11 studies, nine reported that well-being was a potential protective factor against PSMU (Aladwani & Almarzouq, 2016; Demircioğlu & Göncu Köse, 2021; Hou et al., 2017b; Kircaburun & Griffiths, 2018; Kircaburun et al., 2020b; Satici & Uysal, 2015; Shan et al., 2021; Süral et al., 2019; Uysal, 2015). One showed a very small effect size (11.1%), seven showed a small one (77.8%) and one was moderate (11.1%).

Specifically, the studies found the following to be protective factors: flourishing (Satici & Uysal, 2015; Uysal, 2015), self-esteem (Aladwani & Almarzouq, 2016; Demircioğlu & Göncu Köse, 2021), life satisfaction (Satici & Uysal, 2015), social safeness (Uysal, 2015), relationship satisfaction (Demircioğlu & Göncu Köse, 2021), ‘Trait emotional intelligence (TEI)’ (Süral et al., 2019), self-confidence (Kircaburun et al., 2020b), subjective vitality (Satici & Uysal, 2015), self-liking (Kircaburun & Griffiths, 2018), psychological capital (Shan et al., 2021) and psychological resilience (Hou et al., 2017b).

Fear of Missing Out (FOMO). Out of 4 studies, all reported FOMO as a potential risk factor for PSMU (Casale et al., 2018; Dempsey et al., 2019; Sheldon et al., 2021; Varchetta et al., 2020). One showed a small effect size (25%), two moderate (50%) and one a large effect size (25%).

3.3 Internet gaming disorder

A total of nine studies have analysed the predictive factors for IGD in university students.

3.3.1 Description of studies

The design was longitudinal in four studies (44.4%) (Dang et al., 2019; Yang et al., 2021a, 2021b; Yuan et al., 2021; Zhang et al., 2019) and in five it was cross-sectional (55.6%). 77.8% are from Asia (n = 7), one from the USA (Mills & Allen, 2020) and another from Mexico (Borges et al., 2019). 22.2% of the studies were published between 2016—2017, and 77.8% from 2021. The samples ranged from 179 (Kim & Kim, 2017) to 7022 students (Borges et al., 2019), with 88.9% below 1000 and a mean of 1131.9 (SD = 149.5).

The terms ‘Internet gaming disorder’ (n = 7, 77.8%), ‘online game addiction’ (n = 1, 11.1%) and ‘excessive online game usage’ (n = 1, 11.1%) were used.

Eight assessment instruments were identified (see Table 4). High scores indicated a higher degree of IGD. Two studies used the DSM-5 scale (APA, 2013) with a range of 0 to 9 and scores ranging from 1.44 (SD = 1.97) (Zhang et al., 2019) to 1.45 (SD = 1.97) (Dang et al., 2019).

Two studies (22.2%) reported the prevalence. The study by Li et al. (2016), which used the Chinese version of the Online Game Cognitive Addiction Scale (OGCAS; Li et al., 2008), with a range of 16—80 and using a cut-off point of ≥ 32 (plus a score in the ≥ 5 CIA), reported a prevalence of 4.7%. The study by Borges et al., (2019), using an instrument based on the nine symptoms described in DSM-5 and formulated by Petry et al. (2015), with a range of 0—23 and using a cut-off point of ≥ 5, reported a prevalence of 5.2%.

3.3.2 Predictive factors

From the nine studies with adequate quality, a potential predictive factor for IGD was extracted, in the category of psychological variables (see Table 7).

Table 7 Internet gaming disorder predictive factors
Full size table

Psychological variables

Negative affectivity (depression, avoidant coping styles). Out of three studies, all reported negative affectivity as a potential risk factor for IGD (Dang et al., 2019; Li et al., 2016; Yuan et al., 2021). Two showed a small effect size (66.6%) and one a moderate one (33.3%). Specifically, two studies established that depression was a risk factor (Dang et al., 2019; Yuan et al., 2021) while one established avoidant coping style as a risk factor (Li et al., 2016).

3.4 Problematic pornography use

A total of two studies have analysed the predictive factors for problematic internet pornography use (PIPU) in university students.

3.4.1 Description of studies

The design was longitudinal in the study by Grubbs et al. (2018) and cross-sectional in that of Chen et al. (2018). Both were published in 2018. The samples were 808 (Chen et al., 2018) and 1507 (Grubbs et al., 2018).

The terms ‘problematic pornography use’ (Chen et al., 2018) and ‘perceived addiction to Internet pornography’ (Grubbs et al., 2018) were used.

Regarding the assessment instruments (see Table 4), the study by Chen et al. (2018) used the Problematic Pornography Use Scale (Kor et al., 2014) reporting an average score of 7.13 (SD = 8.48) in a score range between 0 and 60 (the higher the score, the higher the degree of problematic use). The study by Grubbs et al. (2018) used the Cyber Pornography Use Inventory‐9 (CPUI‐9) (Short et al., 2012) reporting an average score of 1.7 (0.9) in a score range of 1 to 7 (higher score, higher grade of problematic use). Neither study reported on the prevalence.

3.4.2 Predictive factors

Due to a lack of studies, no predictive factors were extracted from PIPU.

Common and specific factors for problematic Internet and smartphone uses are summarised in Fig. 2.

Fig. 2
figure 2

Predictive factors. Note. PSU: Problematic Smartphone use; PSMU: Problematic social media use; IGD: Internet gaming disorder; FoMO: Fear pf Missing Out

Full size image

4 Discussion and conclusions

Based on previous studies affirming that university students are a population at risk from PIU (Anderson et al., 2017; Ferrante & Venuleo, 2021; Kuss et al., 2014), that PSU and IUP as behaviours overlap in many ways (Carbonell et al., 2018) and that various forms of PIU, including widespread PIU and problematic use of the Internet associated with specific activities (Billieux, 2012; Davis, 2001), the interest in this systematic review has been to complete the study of predictive factors for generalized PIU in this population (Sanchez-Fernandez et al., 2022), focusing in this case on PSU and the specific problem online behaviours that constitute PIU.

As methodological aspects of this systematic review, we can highlight, in the first place, the analysis of international studies with a cross-cultural approach. In addition, the date of bibliographic search, since 2013, is relevant, as this year coincides not only with the date of publication of DSM-5 (APA, 2013), where IGD is acknowledged for the first time, but also with the expansion of smartphones, which make it easier to connect to Internet (Carbonell et al., 2018). Regarding the search strategy, the multiple terms used in the literature to refer to PSU and the specific behaviours within PIU, have been taken into account, which makes it possible to analyse a wide range of studies on these constructs. Finally, with regard to the exhaustive study of predictors, strict criteria have been used in terms of the number of studies that support it, and this furnishes our review with scientific evidence.

With respect to objective 1, to become familiar with the terminology used to refer to PSU and the specific behaviours within PIU, a wide variety of concepts have been found and divided into four groups and subsequently used for the analysis of predictive factors (i.e., PSU, PSMU, IGD and PIPU). Based on the number of articles in each group, research on cyber addictions in university students in recent years has been more focused on the first two (PSU and PSMU). However, no consensus has been found regarding the use of different terms within each group even though there may be differences between them. For example, according to some authors (Kaplan & Haenlein, 2010; Kuss & Griffiths, 2017) ‘social networking’ and ‘social media’ are different concepts even though they are often used interchangeably in literature. The use of ‘social media’ refers to producing and sharing content online, including collaborative projects (e.g., Wikipedia), blogs or microblogs (e.g., Wordpress), content communities (e.g., Flickr), social networking sites (e.g., Instagram) and virtual worlds (e.g., Second Life); while the use of social networks refers to the connection of users (Hamm et al., 2013). This lack of nosological precision has also been reported in other problematic online behaviours such as generalized PIU or PSU (Carbonell et al. (2018), where problematic use and addiction, despite having been established at source as different levels of severity within the same continuum (Young, 1998a; Zhou et al., 2018), are used in research as synonyms.

On the other hand, it should be said that in the case of the PSMU, some studies have analysed the problematic use of specific applications such as “Facebook”, “Instagram” or “WeChat”. However, it is considered advisable to study PSMU in a general way by extending problematic use to a wide range of activities that can take place on social networks, with problematic use of specific social networks such as Facebook being just one example of the PSMU (Kuss & Griffiths, 2017).

In relation to Objective 2, to review the instruments used to assess PSU and specific behaviours of PIU, the instruments that have been highlighted by their frequency are Smartphone Addiction Scale (SAS) (Kwon et al., 2013), in its various forms, in the case of assessing PSU, and the different Bergen scales in assessment of PSMU. However, many different instruments have been found in each of the problematic behaviour groups. Although most are specific to each behaviour, other instruments used are specific to generalized assessment of PIU, such as the IAT (Young, 1998a) in the assessment of PSMU (Hong & Chiu, 2016; Hong et al., 2014; Jaradat & Atyeh, 2017; Kircaburun & Griffiths, 2018) and IGD (Yang et al., 2021a, 2021b). This may be related to the conceptualisation of PIU as an umbrella term that encompasses a number of problematic behaviours on the Internet (Griffiths, 1998, 1999). This heterogeneity and lack of consistency has been confirmed by previous studies that establish the need to develop more advanced assessment instruments in the field of cyber addictions that improve their psychometric properties and allow for a reliable diagnosis (Bányai et al., 2017; Chen & Jiang, 2020; Darvesh et al., 2020; Harris et al., 2020; King et al., 2020; Kuss et al., 2014; Petry et al., 2014; Pontes & Griffiths, 2014; Stevens et al., 2021). A review of Busch and McCarthy (2021) reveals a lack of research to test the functioning of the various PSU measures. In addition, Ryding and Kuss (2020) argue that self-reporting measures are inadequate as we are dealing with unconscious behaviours that are difficult to estimate retrospectively, and therefore propose objective passive monitoring in smartphone research.

In fact, the variety of terminology and assessment tools has affected Objective 3 of this review, which is to analyse the prevalence of PSU and specific PIU behaviours among university students. In the case of PSU, prevalence rates have ranged from 8.99% (Jiang & Shi, 2016), using the Problematic Use of Mobile Phones Scale (PUMPS) (Merlo et al., 2013), to 52.9% (Lian & You, 2017) using the Mobile Phone Addiction Scale Index (MPAI) (Leung, 2008). In the case of PSMU, prevalence rates varied from 14.8% (8.2% ‘‘Excessive use” and 6.6% “serious excessive use”) (Hou et al., 2017a) using the ‘Excessive WeChat Use scale’, to 70% (62.1% with alert levels and 7.9% with addiction) (Jaradat & Atyeh, 2017), using the Internet Addiction Test (IAT) (Young, 1998a). Regarding IGD, prevalence rates ranged from 4.7% (Li et al., 2016), using the Online Game Cognitive Addiction Scale (OGCAS; Li et al., 2008), to 5.2% (Borges et al., 2019), using the Petry et al. instrument. (2015).

However, variability in prevalence rates may be due to other factors. In this respect, studies with the same instrument and the same cut-off point (≥ 31 in men and ≥ 33 in women in the Smartphone Addiction Scale Short Version), prevalence rates ranged from 27.92% (Yuchang et al., 2017) to 48.6% (Gündoğmuş et al., 2021). So, these discrepancies could be explained by socio-cultural differences among users at university (Bányai et al., 2017; Lopez-Fernandez et al., 2017).

In spite of the variability found, the prevalence has generally been higher in PSU and PSMU than in IGD, which may be due to the fact that the first two constructs include a greater number of problematic behaviours. However, these results should be interpreted with caution as the percentage of studies that reported prevalence rates was very low. This in turn could be due to a current conceptualisation of PIU based on a dynamic and procedural view, according to which we would be dealing with differences between levels of severity within a continuum from "normal" to pathological (Ferrante & Venuleo, 2021).

With regard to Objective 4, to study the risk and protective factors associated with PSU and specific problem behaviours online in university students, 10 associated with PSU, four associated with PSMU and one associated with IGD were found, categorised into two types of factors: patterns of use and psychological variables. Following the study by Billieux (2012) these can in turn be categorised as being common to different problematic behaviours or specific to each one of them.

Beginning with predictors common to more than one problematic behaviour, in terms of usage patterns, evidence has been found to affirm that social media use increases the risk of PSU and PSMU in university students (Abbasi et al., 2021; Chung et al., 2019; Coban & Gundogmus, 2019; Dempsey et al., 2019; Gao et al., 2021; Hong & Chiu, 2016; Hong et al., 2014; Jasso-Medrano & Lopez-Rosales, 2018; Koç & Turan, 2021; Lin & Chiang, 2017; Rozgonjuk et al., 2018; Salehan & Negahban, 2013; Sayeed et al., 2020; Varchetta et al., 2020; Xie & Karan, 2019). This is in line with studies which affirm that, among PIU-specific behaviours, using social media has a higher risk of becoming problematic (Carbonell et al., 2018). This result is important because academic use of social media has increased in recent years (León-Gómez et al., 2021; Seaman & Tinti-Kane, 2013), making it necessary that the introduction of social networks in the classroom is accompanied by training in healthy use of social media so as not to increase the risk of problematic behaviour among students.

Regarding psychological variables, negative affectivity was a risk factor common to PSU, PSMU and IGD (Alavi et al., 2020; Canale et al., 2021; Choi et al., 2015; Cui et al., 2021; Dang et al., 2019; Dempsey et al., 2019; Elhai et al., 2018a; Elhai et al., 2020c; Enez Darcin et al., 2016; Forster et al., 2021; Gao et al., 2021; He et al., 2020; Hong et al., 2014; Hou et al., 2021; Hou et al., 2019; Jasso-Medrano & Lopez-Rosales, 2018; Khoury et al., 2019; Kim & Koh, 2018; Kim et al., 2017; Kim et al., 2019; Kircaburun et al., 2020b; Kuang-Tsan & Fu-Yuan, 2017; Kuru & Celenk, 2021; Li et al., 2016; Liu et al., 2020; Liu et al., 2021; Long et al., 2016; Matar Boumosleh & Jaalouk, 2017; Punyanunt-Carter et al., 2018; Sayeed et al., 2020; Xiao et al., 2021; Xie & Karan, 2019; Yang et al., 2020a, 2020b; You et al., 2019; Yuan et al., 2021; Zhang et al., 2020a). In fact, the previous review had already found a risk factor for generalized PIU (Sanchez-Fernandez et al., 2022). This outcome is consistent with model of compensatory internet use aetiological models that suggest that these problematic behaviours may reflect maladaptive coping deployed to regulate negative moods or cope with affective disorders (Kardefelt-Winther, 2014; Kardefelt-Winther et al., 2017), and with cognitive behavioural model of pathological Internet use (Davis, 2001). In line with this, well-being was found to be a protective factor for PSU and PSMU (Aladwani & Almarzouq, 2016; Demircioğlu & Göncu Köse, 2021; Elhai et al., 2018b; Hou et al., 2017b; Jiang & Shi, 2016; Kim & Koh, 2018; Kircaburun & Griffiths, 2018; Kircaburun et al., 2020b; Koç & Turan, 2021; Lian & You, 2017; Lian, 2018; Satici & Uysal, 2015; Shan et al., 2021; Sun et al., 2021; Süral et al., 2019; Uysal, 2015; Yuchang et al., 2017; Zhang et al., 2020). However, there may be two-way relationships between negative affectivity and cyber addictions. Thus, the updated person-affect-cognition-execution interaction model (Brand et al., 2019) states that in the early stages of problematic behaviour, relief in negative affective responses would lead to positive reinforcement which in turn would lead to the establishment of problematic behaviour. As the process progresses and control over the use of specific Internet activities decreases, negative affectivity may be exacerbated by repeated use of the chosen online sites/applications, and problematic behaviour is maintained due to compensatory effects.

This is in line with Busch and McCarthy (2021) who find in their review that emotional health problems are a background to, but also a consequence of PSU, suggesting the need to define and determine how these variables relate. On the other hand, the relationships between cyberaddictions and negative affectivity are not static and can be affected by situational circumstances and traumatic events (Chen et al., 2022).

On the other hand, the Fear of Missing Out (FOMO) – defined as anxiety arising from the belief that others may be having rewarding social experiences which you are not included in (Przybylski et al., 2013) – has been found to be a risk factor for PSU and PSMU (Casale et al., 2018; Dempsey et al., 2019; Elhai et al., 2018b, 2020a, 2020b, 2020c; Handa & Ahuja, 2020; LIN et al., 2021; Sheldon et al., 2021; Varchetta et al., 2020; Wolniewicz et al., 2020; Yuan et al., 2021). Kuss and Griffiths (2017) state that FOMO may be part of social media addiction.

Based on these findings, the introduction in universities of actions aimed at promoting appropriate stress coping strategies and, in general, mental health for the prevention of online problem behaviours is proposed. School-based social and emotional learning (SEL) programs are proposed in the literature (Barry et al., 2017; Dowling & Barry, 2020), which are based on the integration of actions aimed at promoting mental health in teaching practices, and are proven to be effective. This would be in line with the current lines of treatment according to which it would be about, on the one hand, having an impact on emotional health and treating concurrent disorders, such as depression or anxiety (Király & Demetrovics, 2021).

Following the specific predictive factors, in terms of patterns of use, it has been found that the use of the smartphone, in time and frequency, is a risk factor for PSU (Alosaimi et al., 2016; Elhai et al., 2018a, 2018c, 2020b; Erdem & Uzun, 2020; Gökçearslan et al., 2016; Hong et al., 2021; Lin et al., 2021; Wolniewicz et al., 2020). In line with this, the amount of time spent online is a predictor of PIU (Sanchez-Fernandez et al., 2022). In fact, in cognitive-behavioural therapy for problematic Internet use, one of the techniques used is usage monitoring with the goal of reducing the amount of time spent online to a degree that no longer interferes with the client's healthy functioning (Király & Demetrovics, 2021). In the university setting, one way to act on this risk factor would be to promote curricular and extracurricular activities that do not involve smartphone use so that university students do not spend so much time online. However, it must be noted that the studies reviewed have considered the variable ‘time of use’ without differentiating the activities carried out on the network. Huang's meta-analysis (2010) establishes that the effect of internet usage time on PIU is moderated by specific activities (e.g., social vs. non-social). It is therefore recommended that these psychometric limitations be solved by studying the effect of time spent on PSU, distinguishing between time spent on different functions (such as academic, work or entertainment).

In addition, a positive effect of process use—defined as smartphone use involving non-social motivations such as news consumption, entertainment, and relaxation—has also been found in PSU (Elhai et al., 2020a; Rozgonjuk & Elhai, 2021; Rozgonjuk et al., 2019). On the other hand, social use -creating and maintaining relationships- has been found to be a risk factor of PSMU (Hou et al., 2017a; Kircaburun et al., 2020a; Raza et al., 2020; Sheldon et al., 2021; Süral et al., 2019). For this reason, it would be advisable for university institutions to favor alternative forms of face-to-face entertainment and the establishment of social relationships among students.

With respect to psychological variables that specifically predict PSU, impulsivity should be noted (Canale et al., 2021; Khoury et al., 2019; Roberts & Pirog III, 2013; Roberts et al., 2015). It has also been confirmed as a risk factor for generalized PIU (Sanchez-Fernandez et al., 2022). This can be explained by aetiological models that argue that these problematic behaviours may reflect impulse control disorders (Kardefelt-Winther et al., 2017; Young, 1998a). In addition, the PSU model developed by Pivetta et al. (2019), suggests that attention impulsivity predicts addictive and antisocial use of the mobile phone. In the same vein, self-control/self-regulation has been found to be a protective factor for PSU (Cebi et al., 2019; Gökçearslan et al., 2016; Jiang & Shi, 2016; Pourrazavi et al., 2014; Yang et al., 2019, 2020a, 2020b; Zhang et al., 2021), which could be explained by the Larose model of self-regulation (2003), so that people with good levels of self-regulation would be able to activate self-conscious processes that would allow them to judge, monitor and adjust their behaviour online. In fact, Billieux (2012) describes an integrative model in the origin of PSU that includes the impulsive pathway among the different ones. This pathway describes those individuals whose mobile use is motivated by poor self-control and/or poor regulation of emotions. Therefore, preventive strategies that promote self-control and emotional regulation among students could be implemented. For example, mindfulness-based stress reduction (MBSR) intervention has positive effects on the mental health of college students (Canby et al., 2015).

Also, neuroticism/emotional stability is a personality trait that predicts PSU (De Pasquale et al., 2019; Erdem & Uzun, 2020; Liu et al., 2020; Roberts et al., 2015; Takao, 2014). This finding is consistent with the Pivetta et al. model (2019), which establishes a positive relationship between neuroticism and addictive smartphone use. This can be understood through the pathway of excessive reaffirmation (Billieux et al., 2015a) according to which inappropriate use of the mobile phone would be explained by the perceived need to maintain interpersonal relationships and to be constantly encouraged by others.

Finally, other cyber addictions have been found to be a risk factor for PSU (Choi et al., 2015; Khoury et al., 2019; Yuan et al., 2021), which could be explained by another of the pathways of the Billieux (2012) model: the pathway of cyber addiction. Smartphones allow access to the internet and various online activities and so people who make dysfunctional use of the internet or some of these activities would be more susceptible to misuse their smartphone. In fact, this result is in line with the predictors common to the different behaviours found in this review and in the previous review (Sanchez-Fernandez et al., 2022). As a result, there could be mechanisms that would explain the entire spectrum of cyber addictions such as negative affectivity, and it would be very important to promote it from universities.

4.1 Practical implications

This systematic review allow to achieve greater knowledge about PSU and about specific problematic uses of Internet in university students and their predictive factors. The findings may be useful in the development of preventive educational strategies that, implemented from early stages, such as in primary and secondary education, and continuously as lifelong learning, could reduce the occurrence of these online problem behaviours in the university stage. In this way, we are not only contributing to the research needs in this area proposed by the WHO (2015) but also to current European educational policies that aim to support a sustainable and effective adaptation of education and training systems to the digital age (European Commission, 2020). In this way, the ultimate aim is for students to make use of the internet to enable them to be active citizens in the present knowledge society and, at the same time, to minimize the negative repercussions of the network on physical, psychological and social health.

4.2 Limitations and future research

Regarding the limitations of the studies included in this review, the number of studies found that met the inclusion criteria and focused on IGD and PIPU was small, and there was a lack of online gambling studies, which has made it impossible to obtain predictors of these behaviours. In addition, it should be noted that most of the studies included have not met the quality criterion with respect to control of extraneous variables. This means the results obtained have to be treated with caution.

On the other hand, with respect to the limitations of the review itself, in the first place, only the direct effects of the potential predictor variables of problem behaviours have been studied, without taking into account variables that had an indirect effect on the behaviours studied. In future studies, we recommend paying attention to the variables that indirectly influence these behaviours. Secondly, because most of the studies included had a cross-sectional design, with the data extracted it is impossible to determine the causal nature of the predictive factors studied. Thirdly, a very high percentage of studies were conducted in Asia, which may skew the results when generalising for other regions. For this reason, the proposal is to replicate this study in other continents so that the results can be made specific and compared. And fourthly, although the most relevant databases have been used in the topic studied, there could be an information bias that might make it necessary to extend the study using databases not taken into account in this paper.

This study continues the line of research consisting in the systematization of common and specific predictive factors of the different problematic online behaviours, providing a new approach focused on the university population. Based on the findings of this review, further research is needed on predictors of problematic online behaviour such as IGD, PIPU and problem online gambling in this population. Also, future research should use longitudinal designs to establish causal relationships between predictive variables and PSU/sPIU. Regarding recommendations for future prevention programs, these should target to the development of adaptive coping strategies that allow an adequate response to negative emotional states, working in parallel with usage patterns that increase the risk of inappropriate internet use.

5 Conclusions

In summary, this systematic review makes it possible, on the one hand, to reaffirm the need to continue moving forward with the conceptualisation and assessment of PSU and specific PIU in order to achieve consistent diagnostic criteria that, in turn, make it possible to establish the prevalence of these problems in the population in general and in university students in particular. On the other hand, with regard to the predictors reported, our results support the updated version of the I-PACE model (Brand et al., 2019), with the result that the patterns of use and the psychological variables observed increase the prior vulnerability of specific online problematic behaviours by acting as predisposing factors. These behaviours, by interacting with affective and cognitive responses to stimuli, deficits in executive functioning, decision-making behaviour that leads to use of certain applications/Internet sites and the consequences of use of applications/Internet sites, would lead to the development and perpetuation of problem behaviour online. Our findings have made it possible to make progress in the investigation of shared and specific predictive factors of problem behaviours online, thus allowing the formulation of preventive strategies aimed at each one of them.