Occupational risk factors associated with Covid-19 among Health Workers in a tertiary referral Hospital, Nairobi Kenya: A case control study

Conceptual framework

The occupation of a health worker, defined by their clinical roles, is a known risk factor for contracting diseases related to respiratory pathogens such as SARS viruses.¹¹ Schulte et al.’s framework on personal and occupational risk factors to disease outcomes was adapted to hypothesize the relationship between potential Covid-19 exposure risk factors based on the review of literature and the Covid-19 disease status among HWs as shown in Figure 1 below. Other factors that potentially confound the occupational exposure factors and the Covid-19 disease outcome among HWs were considered to guide the variables of measurement for the study (Table 7).

Figure 1.   Conceptual framework on risk factors to Covid-19 among HWs - Adapted from Schulte et al. Model on the association between Personal and Occupational Risk Factors and Disease Outcomes.¹²

As per the conceptual framework in Figure 1, suboptimal PPE and IPC adherence use was hypothesized as the primary exposure of interest which together with other risk factors such as improper PPE reuse, perceived sufficiency of PPE supply and the fit of face masks have been implicated in literature as sources of occupational and nosocomial Covid-19 exposure to HWs.^30–32 Proper use of PPE is an important for the safety of HWs during an outbreak of highly infectious respiratory pathogens such as SARSCoV, MERSCoV, and Covid-19 disease-causing SARSCoV2 virus.³³ Optimal PPE use and compliance rank at the base of the Hierarchy of controls for infectious diseases in the health care setting^34,35 which means that optimal PPE use protects HWs when all other levels of controls fail to guard workers against occupational viral pathogens such as SARSCoV2 virus.³⁶ A systemic review grouped risk factors to Covid-19 disease among HWs into lack of PPE, exposure to infected patients, work overload, poor adherence to infection control measures and pre-existing medical conditions.³⁷ The global shortage of PPE as well as atypical clinical SARSCoV infection presentation were previously cited to pronounce occupational exposure sources of Covid-19 disease among HWs³⁸ as documented in previous SARSCoV outbreak.³⁹ Adherence to face masks, by both the infected and infectious person as the source and the HW as a susceptible host has been underscored in literature as additively protective against Covid-19 disease.⁴⁰ Unit designation or the department of work or Covid-19 designation, has also been cited as a significant contributor to Covid-19 positivity among the HWs in addition to the long duration of duty hours and sub-optimal hand hygiene.^41,42

Developing countries such as Kenya reportedly experienced severe shortage of PPE supplies such as face masks⁴³ during the acute phase of the pandemic. However, knowledge on the effect of PPE shortage on Kenyan HWs remains scant and PPE shortage coping mechanisms such as reuse of PPE has not been scientifically studied or documented based on the Kenyan hospital settings hence its inclusion in the conceptual framework.⁴⁴ PPE reuse has been cited to offer inferior protection for SARSCoV viruses although it has also been cited that reused masks are better than none at all.³² Another literature cites that PPE reuse is a feasible and safe strategy that can be used in helping to slow the spread of Covid-19 disease but this has to be done under specific sterilization autoclaving conditions.⁴⁵ Evidence supporting the decontamination of face masks for reuse cite that use of ultraviolet germicidal irradiation and vaporized hydrogen peroxide is most advantageous.⁴⁶

In line with the WHO recommendations on the need to provide HWs with PPE according to their risk profiles, clinical and non-clinical hospital staff categories have been utilized in health planning on pandemic response and research.^47,48 The current study adopted a broader definition of HWs that includes clinical and non-clinical staff. HWs were thus defined as people employed and working in the Kenyatta National Hospital within the two-month study period whose primary role was the provision of health care services either directly to the patients or indirectly through assistive roles to HWs in line with Kenya Human Resource for Health (HRH) norms and standards 2014 report which enlists staff, clinical and non-clinical workers, required for delivery of health.⁴⁹ Consequently, direct care providers and clinical staff (nurses, midwives, pharmacists, medical doctors and specialists) and indirect care providers, non-clinical and supportive staff (cleaners, porters, security, and departmental administrative staffs such as unit and hospital managers, health information officers and hospital clerks) were studied. This contrasts the narrower definition of health workers as professionals trained in approved institutions and licensed to practice by cadre-specific and respective regulatory bodies to deliver health care following the Kenya Health Workforce Report 2015⁵⁰ which would exclude workers in supportive, unregulated and unlicensed hospital roles.

Previously, HWs engaged in night shift work (NSW) schedules were found to have a higher prevalence of common cold, flu-like illnesses, and gastroenteritis compared to day shift workers.^51,52 Literature exists on various mechanisms by which shift work weakens worker immunity states through a reduction in immune cells such as natural killer cells, CD16, and Inter-luekin 2 factors which are responsible for fighting viral pathogens. In addition, NSW is associated with abnormal cytokine levels.⁵³ NSW is also a direct modulator of the physiology of the immune systems and through circadian rhythm deregulation, it has an indirect effect on individual health status. For example, a systematic and meta-analytic review by Liu et al. 2018 established that night shift work increases the workers’ risk of being overweight and obese.⁵⁴ Both being overweight and obese have been strongly associated with Covid-19 positivity and severe forms of Covid-19 pneumonia.^55,56 Further evidence links sleep habits and the risk of Covid-19 disease among HWs.^57,58 Knowledge on the role of shift work, NSW status and HWs’ sleep habits as occupational and personal risk factors to Covid-19 disease remains incomplete.

The role of staff to staff Covid-19 exposure source among HWs is emerging as observed and reported in a cross-sectional study examining characteristics of HWs with Covid-19 disease in a tertiary public health hospital.⁵⁹ Authors observe that non-significant HW’s occupational type, the peaking of Covid-19 cases among HWs before the peak among Covid-19 disease patients and admissions, and few clustering of Covid-19 among the HWs within certain departments, were suggestive of the predominance of HW to HW spread of the disease rather than the patient to HW spread.⁵⁹ Recent studies on Covid-19 among HWs have also observed that most HWs contracted Covid-19 disease from a non-index case that further supports the predominance of the Covid-19 transmission through staff-to-staff, and surface contamination methods.⁵⁷ The HW to HW spread of Covid-19 may be propagated by common tea and dining rooms breaks where HWs socialize and take meals together while lowering their guard against Covid-19 disease. Hence, more knowledge is needed on common rooms as sources of Covid-19 exposure in occupational settings.

Insufficient HWs’ training on IPC and PPE use has also been implicated in the spread of Covid-19 disease among HWs.^60,61 Less than two hours of training or no training on IPC was associated with a SARS infection among HWs as reported in previous SARSCoV outbreaks.³⁶

This review of literature informed the variables of interest and their relationships as illustrated in the conceptual framework (Figure 1) which further informed the methods of the study in assessing occupational exposure for Covid-19 within the Kenyan hospital setting as detailed in the next section.

Study design and setting

A case-control study design was used. The study was conducted at Kenyatta National Hospital (KNH), the largest public tertiary referral and teaching hospital in Kenya with over 6,000 workers and a bed capacity of 1,800.⁶² The hospital offers specialized health services through its various specialties including accident and emergency unit, specialized outpatient clinics, medical and surgical inpatient units, infectious disease units.

Study population and eligibility of participants

The study population were the health workers (HWs) in full time employment at KNH who were involved in the direct or indirect care of patients with Covid-19 disease or patients with unknown status of Covid-19 disease during the study period between November 2021 and December 2021. This study was conducted just after the peak of the fourth wave of Covid-19 outbreak experienced in Kenya based on the updates from the MOH Kenya.⁶³ Covid-19 testing records at the Covid-19 Testing Unit of KNH were reviewed to identify the confirmed cases of Covid-19 disease among HWs employed and working at Kenyatta National Hospital within the 2 months study period.

Sample size and sampling methods

Sample size determination was done by application of Kelsey's (1996) sample size formula for case control study design as follows:

Of the 160 (40 cases and 120 controls) HWs desired from the sample size formula, a total sample of 147 health workers (39 cases and 108 controls) were recruited. 39 cases with Covid-19 diagnoses were selected through simple random sampling from the RT-PCR laboratory testing records between November and December 2021 and invited to participate in the study. To select controls, matching information on Covid-19 testing dates and cadre for cases was used to identify and conveniently invite 3 potential controls from the Covid-19 database and as referrals from each case in the study as shown in the sampling procedure diagram (Figure 2). Further screening of the HWs in the control group for eligibility was done by use of a screening tool (appendix 6.2) that examined whether a HW had 3 or more symptoms related to Covid-19 and if they had any imaging findings or medical reports suggestive of Covid-19 disease before their negative PCR Covid-19 test results within the study period.

Figure 2. The sampling procedure and the process of enrolling study participants.

Case and control definition

Health workers with a laboratory-confirmed Covid-19 disease, based on RT-PCR SARSCoV2 testing results as per the MOH Kenya case definition for Covid-19 disease, were selected as the cases for the study. Controls were defined as a health workers exposed to the same department of work as a case, PCR tested for Covid-19 but not classified as a suspected or probable, or confirmed Covid-19 case within the last 2 weeks before the Covid-19 test results in the study period between November 2021 and December 2021. The inclusion criteria for controls entailed a HW who reported not having experienced 3 or more symptoms suggestive of Covid-19 pneumonia in the last 2 weeks before Covid-19 test results (either fever and cough of acute onset or any three of these symptoms: Fever, difficulty in breathing, general weakness/fatigue, headache, myalgia, sore throat, coryza, dyspnea, anorexia/nausea/vomiting, diarrhea, altered mental status), and the HW was not classified as either a suspected, probable, or confirmed Covid-19 case in the last two weeks before receiving Covid-19 test results (tested for Covid-19 in the study period with negative test results for SARSCoV2 virus, and did not have any chest X-ray or High-resolution CT scan imaging reports suggestive of Covid-19 pneumonia or disease).⁶⁴ The exclusion criteria for controls included any HW not employed and not working at KNH within the study period, reporting having had either any positive PCR Covid-19 test results within the study period, or had experienced three or more symptoms related to Covid-19 symptoms as per MOH case definition (fever, breathing difficulty, cough, headache, and chills) or had imaging findings suggestive of Covid-19 pneumonia in the last 2 weeks before Covid-19 testing. The screening tool for the study (Table 4) controls was applied to restrict the inclusion of Covid-19 asymptomatic HWs during PCR testing as study controls and exclude symptomatic HWs to minimize misclassification bias for selection of controls (who may have had sysmtomatic and with false negative SARSCoV2 test results).

Data collection

Study participants were interviewed via phone, using a structured questionnaire adapted from the WHO Interim guidance (2020) on Risk assessment and management of exposure of health care workers in the context of Covid-19,⁶⁵ Fatima et al. (2021) study tool to assess night shift status and its association with Covid-19 Infection among HWs⁵⁸ and a case-control study tool by Celebi et al. (2020) utilized to assess the spread of Covid-19 among HWs in common dining rooms.⁶⁶ The questionnaire can be found under the section on Extended data.⁸⁵

Key variables in the study questionnaire on occupational factors included self-reported frequency of PPE adherence, and usage of each PPE including gloves, N95 masks, surgical masks, gown, surface decontamination practice, face shield, moments of hand hygiene measures, and history of the accidental splash of body fluids 2 weeks before PCR testing and attributable type of exposure to Covid-19 case or suspect case 2 weeks before Covid-19 PCR testing, the performance of aerosol-generating procedures 2 weeks before Covid-19 PCR testing, specific department of work within 1 month before Covid-19 PCR testing and self-reported adherence to PPE and IPC protocol in staff dining room in the last weeks before Covid-19 PCR testing (Table 7).

Ethical considerations

Ethical study approval was obtained from Kenyatta National Hospital-University of Nairobi (KNH-UON) Ethics and Research Committee (P462/06/2021). Administrative permission was obtained from the KNH hospital to access laboratory Covid-19 testing records. Physical contacts were avoided to minimize chances of Covid-19 transmission, therefore, interviewer-administered written informed consent was sought virtually and consent recorded on a printed informed consent form and later de-identified once the interview process was over. Identifying contact information of the participants were password protected and used only for the purposes of the current study.

Data analysis

Data was entered into an excel spreadsheet. After double checking for errors, completeness, and accuracy of the data, the excel sheet was exported into R studio version 4.1.2 (2021-11-01) for statistical analysis. Descriptive analysis of the data was reported in form of counts, percentages, and proportions for categorical variables. Categorical variables were derived from the primary data to form aggregate variables including overall PPE/IPC adherence, HW to HW IPC adherence in common rooms, HWs’ clinical status, and departmental risk type categories.

Clinical and non-clinical staff categories were derived to cover HWs who were directly and indirectly involved in provision of care to the patients in the hospital (Table 6). These categories capture all the hospital workers as per the norms and standards for HRH in Kenya.⁴⁹ Departmental risk categories namely, high risk, medium risk, and low-risk- were adapted from the U.S. Occupational Safety and Health Act (OSHA) Guidance on Preparing Workplaces for Covid-19 and the recent studies assessing Covid-19 exposure risk factors among HWs^59,30,67 (Table 5). For this study, the definition for high-risk departments were defined as hospital units that routinely perform high-risk procedures that generate aerosols such as dental units, intensive care units and high dependency unit, accident and emergency units, and any Covid-19 designated units such as infectious disease units (IDU), Covid-19 isolation wards and Covid-19 designated medical wards.³⁰ Medium risk departments were defined as hospital units where staff have frequent contact with patients without confirmed or suspected Covid-19 disease and those with unknown and unspecified Covid-19 status including outpatient medical and surgical clinics, general medical and surgical wards, radiology departments, and maternity. Low-risk departments were defined as units without any contact with patients regardless of Covid-19 status and have minimal contact with patients’ immediate environment such as the administrative and catering departments.

Overall adherence to optimal PPE and IPC measures was assessed using the 13 item criteria for PPE and IPC measures (Table 8). This is supported by hospital guidelines which do not mandate PPE adherence and usage the same way for a clinical staff and for a non-clinical staff especially when PPE supply is constrained.⁵ Therefore, criteria for adherence was applied to clinical and non-clinical staff differently whereby failure to check all the category-specific PPE requirements was classified as either overall sub-optimal adherence or overall optimal adherence.⁶⁵ Frequency of specific PPE use was assessed in four levels, namely, always as recommended (>90% of the time), mostly (>50% of the time), occasionally (20% to under 50% of the time), and rarely (<20% of the time).⁶⁸ HWs who reported having regularly practiced social distancing in the shared dining rooms and having completely doffed before using shared dining rooms in the last 2 weeks before the Covid-19 test conducted within the study period, were categorized to have adhered to IPC in staff dining or common rooms.

The odds ratio OR were reported at 95% confidence intervals and at an alpha of 0.05 (p-value) for statistical significance. Wald’s Odds ratio test for the bivariate analysis in the cross-tabulation of the independent and dependent variables was applied to compute the p-value for variables whose observations had no more than 20% of the counts with less than 5 counts. Otherwise, Fisher’s exact test was applied. Shapiro Wilk test was used to test for normality of the continuous variable (age) in the data set. Under the bivariable analysis, each explanatory study variable was cross-tabulated with the Covid-19 disease status of the HW and the corresponding Odds Ratio (OR) at 95% Confidence Intervals were reported and tabulated.

In the multivariable analysis, a liberal p-value of less than 0.25 was chosen as the cut-off for selecting multiple variables to enter into the multivariable logistic regression model to control for confounding.⁶⁹ Additional independent variables hypothesized to be important confounders in the association between the overall sub-optimal PPE/IPC adherence and the Covid-19 status were added a priori in the multivariable analysis even if they did not attain the 0.25 cut-off for liberal variable selection due to their clinical importance in occupational disease transmission as per the review of literature. These included the sex, overall adherence to optimal PPE/IPC use, department risk type, and the clinical status category of the health worker. The Akaike Information Criteria (AIC) was used to run the modeling simulations of the logistic regression analysis to arrive at the most statistically significant combination of the independent variables that majorly accounted for the variation in the dependent variable among the participant HWs.

Additional hierarchical models to re-introduce variables de-selected by AIC simulation were applied to assess for their impact in the overall model. Likelihood ratio tests were applied for each model to determine the model with the best fit and the model with the least residual deviance which was taken as the most accurate and best fit for the data. A p-value of less than 0.05 was taken to indicate that the fitted model was significantly different from the null model. The effect size was computed by the use of the pseudo R² estimates by the Hosmer and Lemeshow, NegelKerke, and McFadden pseudo R² estimation methods. Interaction terms were also fitted under hierarchical models to assess for significant interaction (p<0.05). An interaction term was applied to assess the effect of gender on the association between department risk type and Covid-19 disease status based on previous evidence of interaction between gender and the effect of department type and Covid-19 disease status.³⁰ Variance Inflation Factor (VIF) was applied to detect any multi-collinearity between the explanatory variables in the fitted model with a VIF of less than 4 indicating a lack of collinearity among the explanatory variables in the model.⁷⁰

Socio demographic and clinical characteristics of the study participants

A total of 147 health workers (39 cases and 108 controls) consented to participate in the study. Overall, 55% of the participants were female with a median age of 35 years (cases) and 37 years (controls) respectively. A slight majority (51%) had professional work experience greater than 10 years, with most (71%) being clinical health providers with direct patient interaction. A significant majority were fully or partially vaccinated against Covid-19 disease (34/39 cases) and (102/108 controls). By co-morbidity status, 74% of the study participants did not have any preexisting medical condition, while 26% reported having a pre-existing condition including hypertension asthma, diabetes mellitus, heart disease, obesity, chronic kidney disease, and cancer (Table 1). The full dataset can be found under Underlying data.⁸⁵

Association between the occupational risk factors and Covid-19 disease status of the HW from the bivariable analysis

Self-reported adherence to PPE including gloves, face masks and N95 mask or equivalents, gowns, face shields and hand hygiene measures was assessed. The crude odds ratio of contracting Covid-19 were 2.57 times higher for HWs who self-reported not adhering to gloves as recommended (cOR, 1.45; 95%CI, 1.11-5.67; p<0.05), 3.5 times higher for not adhering to gown as recommended (cOR, 3.50; 95%CI, 1.35-9.06; p<0.05) and 3.5 times higher for not adhering to face shields as recommended (cOR, 3.50; 95%CI, 1.37-8.36; p<0.05). Self-reported reuse of the PPE within-shift (cOR, 1.86; 95%CI, 0.89-3.90; p>0.05) and between shifts (cOR, 3.00; 95%CI, 0.70-12.50; p>0.05) were not significantly associated with increased risk of Covid-19 disease.

The WHO five moments of hand hygiene including hand hygiene before and after touching a patient (cOR,1.103,95%CI 0.48-3.65, p>0.05), hand hygiene before and after a procedure (cOR,0.88;95%CI 0.41-1.92; p>0.05), hand hygiene after exposure to body fluids (cOR,1.14;95%CI 0.45-2.84, p>0.05) and hand hygiene after touching patients’ surroundings (cOR,0.70;95% CI 0.30-1.54; p>0.05) did not differ between the cases and the controls. There was no difference in the use of surgical or medical masks (cOR, 1.72; 95%CI 0.39-7.50; p>0.05) and N95 masks (cOR, 1.23; 95%CI 0.50-3.00; p>0.05) between cases and controls. Self-reported practice of surface decontamination as recommended did not differ between cases and controls (cOR, 1.87; 95%CI 0.90-4.22; p>0.05) (Table 2).

In the multivariable analysis, overall suboptimal PPE/IPC adherence variable (primary exposure of interest) was excluded by the AIC model simulations. To eliminate collinearity effects, the independent variable on suboptimal PPE/IPC use was not reintroduced in the model owing to its relatedness with specific PPE items such as face shields and gloves. None of the socio-demographic characteristics had any statistically significant association with the Covid-19 disease status of the HW. However, the effect of sex of the HW was impactful in the overall model as a confounder whose statistical control improved the accuracy of the overall model (Table 3). Occupational factors identified to be associated with Covid-19 disease after controlling for covariates in model 1 of the AIC simulation included lack of optimal use of gloves (aOR,4.80; 95%CI, 1.28-20.10; p<0.05), face shield (aOR,3.88; 95%CI, 1.28-13.37; p<0.05) and non-adherence to IPC protocol while using tea-breakout rooms (aOR,7.65; 95%CI 1.54-66.51; p<0.05), working in a medium risk department (aOR,4.38; 95%CI,1.22-18.40; p<0.05) in the last 2 weeks before PCR Covid-19 testing (Table 5). After re-introducing the clinical status in the model to control for the differences in clinical and non-clinical staff categories, statistical significance of suboptimal use of gloves was lost (p>0.05) as shown in model 2 (Table 3).

There was negligible collinearity (VIF <2) between the predictors in the final multivariable logistic model (model 2 in Table 3). In addition, the final model had significant goodness of fit (p<0.05) and had an explicative value estimate of 36.5% (Hosmer-Lemeshow, pseudo R²), 34.5% (Cox and Snell pseudo-R²) 37% (McFadden pseudo R²), and 50.3% (NagelKerke pseudo R²) on the variation in the outcome (Table 3).

Recommendations

Hospitals should reinforce and facilitate optimal adherence to the recommended PPE and IPC measures to prevent occupational infectious disease transmission among HWs especially during infectious disease outbreaks. This should include supporting and reinforcing adherence to facial mucosal protective gears such as face shields among HWs as recommended for SARSCoV and similar highly infectious outbreaks together with other PPE. Institutional measures focusing on IPC around HWs are needed to minimize exposure of infectious disease from staff to staff transmission when sharing common areas such as dining and rest rooms. Future research on occupational infectious disease risk factors should support the validation of exposure risk assessment tool across different health settings for improved external validity of policy recommendations and guidelines around the protection of HWs at all levels.