The study was conducted at the University of Nairobi (UoN), College of Health Sciences (CHS) – one of the six constituent colleges of the UoN. Notably, UoN is the largest higher learning institution in Kenya, whose working conditions closely mimic those of other public tertiary institutions in the country. The CHS consists of five schools (Medicine [SOM], Dental Sciences [SDS], Pharmacy [SOPharm], Nursing Sciences [SON], and Public Health [SPH]) and four institutes (Tropical and Infectious Diseases [UNITID], Kenya AIDS Vaccine Initiative [KAVI], East African Kidney Institute [EAKI] and the Centre for HIV Prevention and Research [CHIVPR]).

An analytical cross-sectional study design was employed to estimate the prevalence and investigate the risk factors for LBP among the college teaching staff of UoN from June 2016 to May 2017. The study was reported as per the STROBE guidelines for reporting observational studies ²⁰ .

The study population consisted of all teaching staff of the CHS eligible to participate in the study. To be eligible for participation, a staff member had to have been employed for at least 12 months prior to the commencement date of the study (May 2017) and have given informed written consent for participation. Moreover, those having LBP due to trauma, infection or tumour were excluded from the study. The sampling frame of teaching staff was secured from the college registry. To obtain the study sample, a stratified random sampling technique (with strata being the constituent Schools and Institutes of the CHS) was used. Within each stratum, a simple random sample was selected, such that the number sampled per stratum was proportional to the size of the stratum. Arguably, stratified random sampling ensures that all strata are represented in the sample and further improves precision of the estimates by removing the between-strata variation ²¹ . A flow chart of the sampling strategy is shown in Figure 1.

An LBP case was defined as a staff member who had a history of pain localised in the lower back (as previously defined) lasting for at least 24 hours within the 12-month study period, had been physically examined by a physician at a health facility and further undergone a diagnostic imaging examination revealing lumbar disc degeneration. Contrastingly, a non-case was a study participant without a previous history of LBP within the same study period.

The required sample size was determined as specified by Kelsey, JL et al. ²² for cross-sectional studies:

n 1 = ( Z α + Z β ) 2 p ¯ q ¯ ( r + 1 ) r ( p 1 − p 2 ) 2 q ¯ = 1 − p ¯ n 2 = r n 1

p 1 = p 2 O R 1 + p 2 ( O R − 1 )

p ¯ = p 1 + r p 2 r + 1

Where:

n ₁ is the number of cases and n ₂ is the number of non-cases; p ₁ is the proportion of individuals who did not exercise and had LBP; p ₂ is the proportion of individuals who exercised and had LBP – estimated to be 43.1% based on a previous study ¹⁴ . Notably, Z _{α /2} (1.96) and Z _β (-0.84) are the values which specify the desired 2-tailed confidence level (95%) and statistical power (80%) respectively. The odds ratio ( OR) for the effect of the primary exposure (lack of physical exercise) was hypothesised to be 2.2 ¹⁴ . The ratio ( r) of unexposed to exposed individuals was set at 1. Given these figures, a total sample size of 204 participants was derived.

Initially, two research assistants were recruited and trained to aid with the data collection exercise that spanned a two-month period May 31 ^st–July 31 ^st 2017. As for the data collection, a semi-structured questionnaire (see extended data ²³ ) was administered to the study participants capturing details of their LBP history, and predictors: work-related (length of working day, office chair design, stress, social support and job satisfaction) and socio-demographic characteristics (age, sex, marital status, body mass index (BMI), level of education, school (including department), physical activity engagement, tobacco use and level of alcohol intake). The predictors were assessed as given in Table 1. A conceptual framework depicting the predictor-outcome relationship is displayed in Figure 2.

The study respondents provided written informed consent expressing their willingness to take part in the study. Approval for the study was granted by the Kenyatta National Hospital and University of Nairobi joint Ethics and Research Committee (KNH-ERC/A/171).

Granted that cross-sectional studies are prone to a range of biases that may invalidate study results, deliberate attempts were made to minimize their occurrence. Ergonomic and lifestyle factors are readily modified once individuals are diagnosed with LBP. As such, to reduce the possibility of reverse causality involving these set of factors, specific questions targeted the period preceding the onset of symptoms characteristic of LBP for case respondents. To standardise the interview process and thus minimize interviewer bias, the research assistants were trained on sound interviewing techniques. As non-response may introduce selection bias in cross-sectional studies, non-responders were aggressively followed up with reminders to achieve a reasonable response rate.

Prior to data entry, questionnaire responses capturing qualitative variables were coded. The data were then double-entered in an EpiData v3.1 spreadsheet by two independent data entry clerks to minimize errors. The validated dataset was then exported to Stata v13 software for data cleaning and analyses. Continuous variables were summarized using the median and inter-quartile range (IQR) as well as histograms and boxplots. For categorical variables, proportions were computed. The prevalence of LBP and its associated 95% exact binomial confidence interval were estimated. Code for analysis is available as extended data ²⁵ .

For univariable analyses, a mixed-effects logistic regression model was used to evaluate the effect of each predictor on LBP, with the variable department included as a random effect to account for clustering of the outcome within departments. The significance of each of the predictors at this stage was evaluated at a liberal P≤0.20. As inclusion of age, BMI and length of working day as continuous predictors in the univariable models yielded insignificant results, these were categorised and reassessed for significance. In particular, age was grouped into three categories: ≤43yrs; 44–57yrs; ≥58yrs, BMI was classified into the four BMI categories ²⁶ : Underweight(<18.5), Normal weight (18.5–24.9), Overweight (25.0–29.9) or Obese (≥ 30.0) and length of working day was categorised into: ≤8hrs or >8hrs.

Variables that were found to be significant in the univariable analyses were then offered to a multivariable model where a backward step-wise approach was used to eliminate variables at P≥0.05. To minimize confounding, elimination of non-significant predictors was only considered when their exclusion from the model did not result in a more than 30% change in the effects of the remaining variables ²¹ . Two-way interactions were fitted between the remaining variables of the final model and assessed for significance.

A total of 204 teaching staff of the CHS were invited to participate in the study, from whom 167 consented to participating in the survey, giving a response rate of 81.9%. However, of the 167 participants, 31 were excluded from the analyses for reporting trauma and/or infection as the reason(s) for their back pain. Therefore, 136 participants were considered in the analyses [see underlying data].

Descriptive statistics for the predictors of LBP are displayed in Table 2. Notably, the median age for the participants was 51 years (Range: 31–81yrs). A typical working day was 10 hours long (range: 4–18hrs). Only 44.9% (n=61) of the participants regularly exercised. Participants with office chairs that had lumbar support represented 41.9% (n=57) of the total. The estimated 12-month period prevalence of LBP was 64% (95% CI: 55.3%–72.0%).

Based on results of the univariable analyses, the variables: sex, age, school, physical exercise, office chair design and level of workplace stress, were significantly associated with LBP at P 0.20 ( Table 3). These were subsequently offered to the multivariable model. In the multivariable analysis, only physical exercise, office chair design and level of workplace stress were shown to be significant predictors of LBP at the 5% of significance level ( Table 4).

Compared to respondents who regularly exercised, participants who rarely and never exercised had respectively about three (aOR: 2.8; 95% CI: 0.9–8.4) and six times (aOR: 6.0; 95% CI: 1.2–29.6) the odds of LBP controlling for their office chair design and workplace stress level. Participants who sat on chairs with lumbar support had a third (aOR: 0.3; 95% CI: 0.1–0.9) the odds of LBP as those who did not regardless of their level of physical activity and stress at their workplace. Irrespective of their level of physical activity and design of their office chair, respondents who experienced high and low stress levels at their workplace had roughly four times (aOR: 4.4; 95% CI: 1.1–17.5) and three-fifths (aOR: 0.6; 95% CI: 0.2–1.9) the odds of LBP respectively, as those whose perception of stress was medium.

The prevalence of LBP among teaching staff of the CHS, UoN was estimated to be 64.0%. This is a higher prevalence than demonstrated by most studies conducted among teachers in which LBP prevalence ranged between 22.3% (Thailand) and 57.5% (Ethiopia) ^{12– 15} . This variation could be attributable to age differences between the study participants, with those in the mentioned studies being on average younger (mean age: 34.7–38yrs) than those included in the present study (mean age: 50.9yrs). An age-LBP association has been demonstrated, with LBP being more prevalent among individuals over 40 years ^{12, 14, 27} .

This study has shown that teachers who either do not exercise or do so infrequently, have higher odds of experiencing LBP than their counterparts who exercise regularly. This finding is consistent with study observations made in Israel, Iran, India, South Korea and Ethiopia ^{11, 14, 15} . Regular physical exercise has been shown to strengthen lower back muscles and maintain the spine in proper alignment for optimal function. Furthermore, routine exercises increase blood supply to the spine muscles, joints and intervertebral discs minimizing injury and enhancing their repair ^{14, 28, 29} . It has been suggested that a minimum of 30 minutes of regular exercise could increase trunk flexibility and stimulate an adequate production of endorphins that could diminish pain sensation ^{30, 31} .

Sitting on a chair with back support had the effect of lowering the odds of LBP. The use of lumbar supports has been widely advocated because of their well-known function of preserving the integrity of the low back curves, thus reducing the risk of LBP ^{32– 34} . Additionally, the tilt of the lumbar support permits the person using it to sit with his/her upper body slightly reclined which ensures proper body weight distribution ^{32, 35, 36} .

There was a noticeable association between perceived stress levels at the workplace and the reporting of LBP. High stress levels have been associated with the stimulation of the sympathetic nervous system prompting the release of stress mediators that can strain the musculoskeletal system resulting in LBP. Our finding concurs with that reported by an Ethiopian study in which participants reporting stress had roughly two times the odds of experiencing LBP than those without stress ¹⁴ . Similar associations have also been observed elsewhere ^{11, 28} .

Our study did not suggest any evidence for the existence of a real difference in LBP prevalence between sexes. This could be partly ascribable to our study participants being generally older and hence exposed to a similar risk. Nevertheless, among younger participants, being female has been associated with an elevated risk of LBP owing to hormonal imbalances ^{11, 14, 37} . More so, during pregnancy, hormonal changes responsible for loosening the spinal ligaments coupled with the extra weight that stresses the lower back muscles heighten the risk of LBP ^{38, 39} .

Taking into account other study variables, age did not emerge as a significant predictor for LBP in the present study. A likely explanation for this would be that older participants aware of their disproportionately higher risk, engaged themselves in regular exercises at a comparably higher frequency (as per the data: P=0.02) thus arguably, balancing out their LBP risk to that of their younger counterparts. Nonetheless, a number of studies have reported age as a significant risk factor for LBP; old age being associated with spine and vertebral disc degeneration as well as loss of connective tissue elasticity that can result in LBP ^{12, 14, 29} .

Working at a particular school did not significantly influence a participant’s likelihood of LBP. This is conceivable considering that the respondents are likely to have similar work responsibilities entailing preparation of teaching materials, lecturing, grading of students’ papers, mentoring and supervision that often demand extended periods of sitting and standing. This prevalence homogeneity may also denote uniformity in the distribution of office furniture designs across schools.

A couple of limitations are inherent in the present study. Definition of the outcome was pegged on self-report which could have introduced non-differential misclassification with a potential to bias the estimated odds ratios towards null. As measurement of exposures relied on recall which could have been incomplete especially for chronic cases, this would have the potential of biasing the effect estimates. Since non-responders tend to systematically differ from responders with regards to a range of health outcomes, it is anticipated that the current study’s prevalence underestimates the true burden of LBP in this study population. In light of the above-mentioned limitations, it should be borne in mind that the results of this study are merely hypothesis generators and hence, stronger study designs such as case-control or cohort studies are recommended to validate the findings.

The raw dataset for the study is kept under restricted access since it contains sensitive participant information. Access to the raw data is possible upon placing a formal request to the corresponding author ( disykgn@gmail.com). The replication data, analysis script and questionnaire for this manuscript are available from figshare.

Figshare: LBP_UoN_CHS_2.dta. https://doi.org/10.6084/m9.figshare.8148779.v1 ⁴⁰

This project contains the following underlying data:

Data are available under the terms of the Creative Commons Zero "No rights reserved" data waiver (CC0 1.0 Public domain dedication).

Figshare: LBP Study Questionnaire_UoN. https://doi.org/10.6084/m9.figshare.8197598.v1 ²³

This project contains the following extended data:

Data are available under the terms of the Creative Commons Zero "No rights reserved" data waiver (CC0 1.0 Public domain dedication).

Figshare: LBP study_CHS_UoN_Stata code. https://doi.org/10.6084/m9.figshare.8197715.v2 ²⁵

This project contains the following extended data:

Data are available under the terms of the MIT licence.