Chronic Neck Pain Prevalence Before and After COVID-19 Restrictions and Its Relationship With Digital Device Screen Viewing: A Population Study

Ethical approval

This study was conducted in accordance with the principles of the Declaration of Helsinki, and the study protocol was approved by an Institutional Committee on Ethics in Research of Universidad de Piura (Act No. 14/2022). All participants were asked for informed consent. Statistical analyses were performed using an anonymized database and reported according to the STROBE guidelines.

Study design

This analytical, cross-sectional study was conducted in the Republic of Peru, a Latin American country divided administratively into 25 territories (grouped into five regions: Lima, North, Center, South, and East). The survey was population-based and conducted between November 24 and 25, 2022. The target population included individuals aged >18 years living in urban and rural areas. According to official data, the population size in 2021 was 24,290,921 individuals (50.2% females), with 79.9% residing in urban areas [22].

Sampling design

The sampling design was computed with a margin of error of 2.83%, a maximum variance of population proportions (p=0.50), and a 95% confidence interval (95% CI), resulting in a calculated sample size of 1,202 individuals. We employed a multistage sampling method. The first stage consisted of a probabilistic sample selection of locations via systematic random sampling, proportional to the number of inhabitants in each location. The second stage involved systematic sampling, with random selection of blocks of houses, where the probability of block selection was proportional to the number of houses. The third stage comprised the selection of houses through systematic sampling, with a random starting point. Finally, in the fourth stage, individuals within each household were selected based on their sex and age, to achieve the required distribution.

To perform the first, second, and third sampling stages, the cartography of the Instituto Nacional de Estadística e Informática del Peru generated during the national census of 2017 was utilized as the sampling frame [23].

Definition of variables

The variable of interest, CNP, was measured using two questions, each targeting different intervals. The first question inquired: “Considering the last 6 months, have you experienced any pain in the neck, nape, or in the tops of your shoulders?” The second question evaluated a similar occurrence preceding the onset of the pandemic: “Did you experience any persistent pain in the neck, nape, or on the tops of your shoulders before March 2020 (prior to the COVID-19 lockdown) persistently, that is, lasting 6 months or more?”

For both questions, respondents were presented with five response options: (1) “Yes, I experienced pain in that area once a month on average;” (2) “Yes, I experienced pain in that area once a week on average;” (3) “Yes, I experienced pain in that area daily or almost daily;” (4) “No, I have not experienced any pain in that area;” and (5) “I could not tell.” Participants who responded to categories 2 and 3 were considered CNP cases.

The exposure variable was the number of hours spent viewing DDCI screens, measured through a third question: “On average, how many hours a day do you spend looking at a screen of a device connected to the internet, such as your cellphone (smartphone), laptop, personal computer, or tablet for study, work, or entertainment purposes?” Respondents were provided with the following response categories: (1) “I do not look at any screen, or I barely do it;” (2) “I spend <4 h/day looking at a screen;” (3) “I spend between 4 and 8 h/day looking at a screen;” (4) “I spend ≥8 h/day;” and (5) “I could not tell.”

The covariates were demographic characteristics, including age (years), sex, and household socioeconomic level (from “A” indicating higher resources to “E” indicating fewer resources). This socioeconomic variable was derived from the income of the household head, goods and services accessed, household appliances, and access to public services, as previously described. Other covariates included were resident area (urban, rural) and region (Lima, North, Central, South, and East).

Survey

Fieldwork was conducted by IPSOS Opinión y Mercados S.A. organization through face-to-face surveys conducted in the selected households. Data were collected by local resident interviewers who have experience in survey applications. Before administering the survey, the interviewer requested each participant’s consent. In instances where the selected participant was not at home, three additional visits were made. However, if the participants could not be reached after these attempts, they were replaced with individuals matching their age and sex characteristics.

Data collection was facilitated through mobile devices using the iField application (IPSOS Group S.A.), an integrated computer-aided personal interview platform. This approach enabled efficient location tracking of interviewers and participants during the survey, real-time quality control of data, reduced errors in data collection, closed card sorting, and ensured adherence to the sample selection process in assigning households.

Statistical analysis

An exploratory analysis confirmed the absence of data loss; thus, we analyzed the complete dataset. Continuous variables are presented using the mean and 95% CI. Categorical variables are expressed as weighted proportions. Additionally, unweighted absolute frequencies were provided, considering that the weights applied enable the expansion of frequencies to estimate the absolute parameter. The weighted proportions provided corrected and valid percentages representative of the population ≥18 years residing in Peru in November 2022.

The prevalence of CNP was estimated along with the corresponding 95% CI for two-time points: before the COVID-19 pandemic and during the return to normalcy (COVID-19 post-restrictions period). For this analysis, the primary outcome was CNP, defined as occurring daily or almost daily (response category 3) or at least once weekly (response category 2) onset. These frequency levels were selected for prevalence estimation, consistent with previous studies [24,25]. Prevalence was calculated by considering the total of participants in the survey as the denominator.

To estimate the dynamics of CNP changes before and after COVID-19 social restrictions, we utilized a contingency table, with CNP categories before restrictions forming the rows and categories for the post-restriction period forming the columns. Changes were depicted as weighted proportions. These prevalences were compared using McNemar’s test.

To evaluate the association between CNP and DDCI screen use, we considered CNP occurrence after COVID-19 social restrictions. Pearson’s chi-square test with a second-order Rao-Scott correction was used to compare the CNP proportions across strata defined by the study variables. This analysis was conducted among a subpopulation of participants who provided valid responses to the questions, excluding those who responded, “I could not tell.” Additionally, factors associated with DDCI screen-viewing were also identified. A subpopulation of participants with valid responses to this question was used for analysis. At this stage, a statistical criterion (p<0.20) was applied to define the variables included in the regression model.

Ordinal logistic regression was employed to evaluate the association between DDCI screen-viewing and CNP. This model was chosen because CNP, the dependent variable, was measured using categories that reflected an order of frequency (i.e., no pain, daily or almost daily pain, weekly pain, and monthly pain). Prior to fitting the model, we evaluated whether the included variables met the proportional odds assumption by conducting the Wald test for parallel lines. A significance level >0.05 implied that the odds were proportional. This analysis was conducted independently for each variable and set of model variables. IBM SPSS Statistics version 25.0 (IBM Corp.) was used to verify these assumptions.

In the multivariate analysis, we formulated a complete model with all covariates at p<0.20. Additionally, socioeconomic status was included as an epidemiological criterion. Subsequently, a second (reduced) model was constructed, incorporating factors that were significant in the first multivariate model alongside the exposure of interest. Adjusted odds ratios (ORs), their corresponding 95% confidence intervals (95% CIs), and the estimated coefficient were presented for both models. McFadden and Nagelkerke’s pseudo-R-squared values were estimated to assess the global goodness-of-fit. Post-hoc statistical power calculations for the regression model were performed, considering factors such as the sample size, the number of predictors included in the models, the observed R-squared value, and a significance level of 5%.

Statistical analysis was performed for a complex survey sample using the svy command in Stata Statistical Software (Release 16; StataCorp LLC). Statistical significance was set at p<0.05.

Changes in the prevalence of CNP

During the return to normalcy, the prevalence of daily or almost daily CNP was 14.8% (95% CI, 12.6–17.3), while the prevalence of CNP occurring at least once a week was 27.8% (95% CI, 24.9–30.9). These findings represented a significant increase compared to the estimates before the restrictions for both outcomes (McNemar’s test, p<0.001).

Post-restriction, the prevalence of daily or almost daily CNP was 2.43 times higher than before the COVID-19 pandemic (95% CI, 1.78–3.08). Similarly, the prevalence of CNP occurring at least once a week increased by 2.25 times (95% CI, 1.86–2.64). Additionally, we estimated the difference in the weighted prevalence between the return to normal and before the restrictions for each population stratum (Table 2).

Table 2.

Prevalence of daily chronic neck pain daily or almost daily, and at least once a week on average for the sample and according to strata

Among participants who did not experience any pain or had pain very rarely before the restrictions, 60.7% remained painless during the return to normal. For those who reported pain once a month before the restrictions, 17.0% and 12.3% reported an increase in frequency to once a week and daily or almost daily pain, respectively, during the return to normalcy. Additionally, among participants who experienced daily or almost daily pain before the restrictions, 45.9% remained at the same frequency of pain during the return to normal (Table 3).

Table 3.

Changes in the recurrence of chronic neck pain before COVID-19 pandemic and during the return to normality

Screen viewing

Among the entire sample (n=1,202), 23.5% (95% CI, 20.7–26.6) of the participants reported that they did not view screens or did so rarely. In contrast, 39.6% (95% CI, 36.4–43.0) reported viewing screens for <4 h/day, while 20.1% (95% CI, 17.7–22.8) reported viewing screens between 4 and <8 h/day and 12.3% (95% CI, 10.1–15.0) declared viewing screens for ≥8 h/day. Only 4.4% of the participants could not specify how much time they spent viewing screens.

Covariates associated with CNP after the COVID-19 restrictions

In bivariate analysis, the proportion of CNP at any frequency was higher among females than males. Age was also significantly associated with CNP, with a higher proportion of daily CNP observed among individuals aged 60 and above. Furthermore, the crude analysis showed an association between CNP and exposure to DDCI screens (Table 4).

Table 4.

Crude analysis of associated factors with chronic neck pain during the return to normality in adults of Peru

Conversely, the number of hours spent viewing screens was associated with age, area of residence, region, and socioeconomic status. Among individuals aged 18 to <30 years, 25.9% reported viewing screens for ≥8 h/day, while only 2.9% of those aged ≥60 years engaged in the same activity. By residence area, 40.1% of participants residing in rural areas reported not viewing screens, whereas this proportion was 21.2% among those living in urban areas. Additionally, individuals with higher socioeconomic status exhibited higher screen usage (Supplementary Table S1).

Ordinal logistic regression

In the first model, the number of hours spent viewing DDCI screens was included as an independent variable, while the covariates comprised sex, age (in years), socioeconomic status, and region. The covariates met the proportional odds assumption in the bivariate analysis, except for age, which had a p-value of 0.006 in contrast to parallel lines. Nevertheless, we applied ordinal logistic regression since the global model met the proportional odds assumption.

Both the complete and reduced models revealed that viewing DDCI screens, sex, and age were independently associated with CNP. In the reduced model, those who viewed screens for ≥8 hours had 61% higher odds of increasing the frequency of CNP from one level to another compared to those who did not view screens or rarely did. Furthermore, females exhibited 2.31 times the odds of increasing the frequency of CNP compared with males, and each additional year of life increased the odds of increasing the frequency of CNP by 2% (Table 5). The post-hoc statistical power reached by the sample was 99.9%, considering the reduced model (three predictors), MacFadden R²=0.0299, a significance level of 5%, and 1,159 observations.

Table 5.

Ordinal logistic regression to evaluate the association between viewing at internet-connected electronic devices’ screen and chronic neck pain in the Peruvian population

Strengths and limitations

Our study has several strengths. First, the sampling design generated prevalence estimates for the nationwide adult population in Peru according to the country areas (urban/rural) and macro-regions. Second, to our knowledge, this is the first study to compare CNP prevalence before and after the COVID-19 pandemic restrictions on a nationwide scale. This is particularly noteworthy as previous studies have primarily estimated this change in specific occupational groups or patients attending healthcare facilities. Third, the survey was conducted in a structured manner using questions previously applied in a population study in 2016 [20], thus allowing for a comparison of CNP prevalence changes over time. Fourth, this study placed emphasis on measuring pain specifically in the cervical region. Consequently, the questions were tailored to target this specific body area, unlike other studies that employed more generalized assessments or evaluated pain across multiple body areas. This approach served to mitigate recall bias.

This study also has some limitations. Firstly, the identification of CNP was based on self-reports rather than case history or physical examination conducted by healthcare professionals. Secondly, estimates of pre-pandemic CNP rates may be subject to recall bias, as participants (interviewed in November 2022) were asked to remember the presence and frequency of neck pain in March 2020 and at least 6 months earlier. However, it is known that the presence of unusual historical or major events (in this case, the beginning of the pandemic lockdown) tends to improve recall. Thirdly, certain population characteristics, such as education, occupation, or history of COVID-19, were not captured to better characterize the distribution of CNP. This limitation also extended to the study of the relationship between CNP and screen-viewing as they were not included as potential confounders. Fourthly, exposure was also measured via self-reports, and establishing a causal association with CNP was challenging due to the inability to define the temporal sequence between exposure and outcome. Lastly, our study did not explore the intensity or specific causes of CNP.

In conclusion, after lifting COVID-19 social restrictions in Peru, CNP occurring daily or almost daily, or at least once a week, affected approximately four out of ten adults, representing over a two-fold increment compared to the pre-pandemic values. The most affected subgroups were females and adults aged >60 years. Additionally, those who spent ≥8 h/day viewing DDCI screens were at increased risk of suffering CNP compared to those who did not, and this association was found to be independent of age and sex. These findings underscore the broader public health challenges stemming from the COVID-19 pandemic. Moreover, our research revealed that viewing DDCI screens for extended periods is a specific and modifiable risk factor, which is more widely spread across the general population and a potential target for developing preventive strategies to reduce the prevalence of CNP and accompanying disabilities.