Malawi, with 19 million inhabitants, has the second highest cervical cancer incidence and mortality in the world with age standardized rate of 75.9 and 49.8 per 100,000, respectively [18, 19]. About 27% of Malawian women receive cervical cancer screening and only 43% with positive results receive care. HIV prevalence among women 15–49 years is 8.8% [20]. In 2021, Malawi had 0.14 doctors and 0.33 nurses and midwives per 1,000 people [21]. This critical shortage of health workers makes it difficult to increase CCSPT services, particularly in rural areas where there are few providers. CHWs (known as health surveillance assistants in Malawi) may help to address shortage challenges as this cadre is increasingly leveraged to address service provision bottlenecks. CHWs are salaried and employed by the Ministry of Health [22]. They provide their services through Outreach and Village Clinics. Ideally, one CHW should cater for 1,000 community members, but as of 2021, the ratio was 0.55 CHW per 1,000 people [21].
This study was done at 7 health facilities in Lilongwe and Zomba, two of Malawi’s 28 districts serving a population of about 3.8 million [18]. The 7 health facilities were selected because they were all randomized to implement a prevention of cervical cancer through a human papillomavirus (HPV)-based screen and treat intervention with a community component involving CHWs.
Cervical cancer screening and prevention therapy
Details of the CCSPT intervention have been provided elsewhere [16]. Briefly, the intervention integrates HPV-based CCSPT into family planning (FP) services with two models of care. In facilities randomized to model 1 (clinic-only), self-sampling for HPV is offered to women 25 to 49 years with no history of total hysterectomy attending FP services at health facilities. Same-day thermal ablation is provided to HPV-positive, ablation-eligible women. In facilities randomized to model 2 (clinic + community), screening is offered as in model 1 and in addition, community-based screening is also offered to eligible women by CHWs. The CHWs transport collected samples for HPV testing to health facilities, collect HPV results and refer HPV-positive women to the nearest health facility for treatment.
Study design
We conducted an observational time and motion study to describe the work activities undertaken by CHWs at the 7 model 2 facilities. Independent data collectors followed CHWs throughout the workday and performed continuous direct observations of both clinical and non-clinical activities in which the providers engaged [23]. We used a pre/post approach to assess the impact of CCSPT on CHWs time use. The CHWs providing community based CSSPT were each observed twice. Baseline data collection occurred from October to November 2019 prior to CCSPT implementation. Endline data collection occurred from July to August 2021, one year following the roll out of the CCSPT intervention, when the health facilities were judged to be in a steady state of routine CCSPT service provision.
Study eligibility
CHWs who had undergone CCSPT training and were providing services in model 2 facility catchment areas were invited to participate. Trainee CHWs were excluded from the study since they did not have Village Clinics and would thus not be accessed by eligible women.
We recruited CHWs who agreed to participate by giving verbal informed consent. Once recruited before the CCSPT implementation, we aimed to observe the same CHWs after the CCSPT implementation. To minimise the Hawthorne effect, or the change in some aspect of the observed behaviour due to awareness of being observed [24], each CHW was observed twice during each period. The first observation session was a simulation, meant to allow the CHWs to habituate to being observed and facilitate more natural behaviours during official data collection. The second session was for actual data collection. We used data from the second sessions only for analyses. CHWs were not aware that only data from the second observations would be used for data analysis. CHWs showed some awareness of being observed during the first observations as they made effort to explain to observers what were about to do/ doing or sometimes attempted to initiate conversations with the observers. In such cases, observers were instructed to politely remind the CHWs to ignore them (observers) and focus on their tasks. Hardly any of such CHWs behaviours were noted during the 2nd observations.
Observers’ training
Ten research assistants served as observers, none of whom were CHWs themselves. They all underwent a two-day classroom-based training in time and motion studies led by an experienced trainer (JC). The specific roles and activities of CHWs in Malawi are outlined in a national guideline [22]. This informed the development of time and motion activities. During the training, the observers studied these main activities and corresponding sub-activities (Appendix 1). They were instructed by an experienced Community Nurse (EZ) to identify the start and end of each sub-activity without the need to ask the observed CHWs what they were doing. The research assistants were also instructed on how to collect data using tablet computers. A one-day pilot session in which the research assistants practiced observations on non-study CHWs was conducted after the classroom-based training. The observers were encouraged to ask questions to the observed or instructors for clarifications during the pilot session. Experiences and lessons learnt from the pilot were used to modify and improve sub-activity definitions and grouping of activities including refining the structure of the data collection tool.
Main activities, sub-activities and analysis groups
A pre-defined tool composed of a set of activities logically organized to facilitate data collection and analysis was used to document CHWs activities (Appendix 1). CHWs tasks were categorised into 7 main activities: 1) Under 5 children treatment services; 2) Under 5 children preventive services; 3) FP services; 4) CCSPT services; 5) Over 5 services; 6) Administration; and 7) Non-work-related tasks. Each of the main activities had corresponding sub-activities (Appendix 1). Because the CCSPT period coincided with the COVID-19 pandemic, the endline data collection tool was modified to include a COVID-19 sub-activity. We structured the sub-activities so that they could easily be visually identified when each started and ended without the need for CHWs to explain what they were doing. This structure was crucial to the success of the data collection because the observer’s role during data collection was passive involving no communications with the observed.
For analysis, the sub-activities were further collapsed into five groups: 1) Clinical and preventive services; 2) FP; 3) CCSPT; 4) administration and 5) non-work-related tasks. Clinical and preventive services included the whole spectrum of curative and preventive services offered to children < 5 years of age, such as history taking, testing and treatment for acute illnesses, provision of immunizations and growth monitoring. This task grouping also included sub-activities provided to those ≥ 5 years such mass administration of deworming drugs to school age children and provision of diagnostic services to adults for infectious diseases such as tuberculosis and HIV including making appropriate referrals to health facilities. FP services included providing health education to women on FP options and delivery of preferred methods. CCSPT services included counselling about self-sampling for HPV, transporting samples to facilities for testing, delivering HPV results to women and making referrals. Administrative tasks consisted of general duties such staff meetings, official travels, supervisions and report writings. The non-work-related task group consisted of all informal activities, including making personal calls, staying idle, chatting with other health workers and sending SMS or WhatsApp texts for personal or non-work-related purposes.
Data collection
We used a structured data collection instrument (Appendix 1) programmed digitally with Open Data Kit software to facilitate electronic data capture using Samsung Galaxy-Tab-2.0 tablets [25]. Each main activity group appeared as a menu. Thus, to log a sub-activity, the observer had to first identify the main activity under which the sub-activity was listed, then select the main activity, and finally select the sub-activity of interest (Fig. 1). A tap on the start and end buttons by the observer initiated the start and end of a sub-activity timing, respectively. The sub-activity durations were automatically timed by internal clock of the tablet computers between tapping of the start and end buttons. The tablet was also used to capture additional information about the CHWs being observed including their age, sex, years of work experience, catchment population and the name of the nearest health facility.
Only one sub-activity at a time could be captured by the tablet computers. In cases where the CHWs were performing more than one task simultaneously (rare events), for example, examining a sick child while talking to a colleague, it was up to the observer to decide which sub-activity was the dominant one to be timed. For some sub-activities, CHWs often switched between tasks or sub-activities, for instance administering vaccines to children, chatting with a colleague, going back to vaccine administration and then chatting on WhatsApp. The tool was flexible enough to accurately capture data for such sequence of fleeting sub-activities. But the tool was unable to collate tasks that were interrupted by other tasks. For example, if a CHW was engaged in history taking with Patient A, then briefly left to provide contraception to Patient B, and finally returned to complete history taking with Patient A, the data collection system would register the provider’s actions as three distinct tasks (i.e., two tasks of history taking and one task of contraception provision). Thus, activities were observed in singularity. The data collection system did not allow for activity concurrency (i.e., multi-tasking). At the end of each day, data were backed up in the computer tablets and uploaded to a central server for quality control and safe storage.
Data analysis
The primary outcome of interest was the average time in hours spent by a CHW providing routine clinical and preventive services during an average day after the CCSPT implementation. The rationale for choosing this outcome was to ascertain effects of CCSPT on traditional CHW curative and preventive services. Secondary outcomes of interest were time in hours spent providing FP, CCSPT, administrative duties and doing non-work-related activities per CHW. The time spent on each outcome was estimated by adding up all sub-activity durations in minutes under its respective analysis group (clinical and preventive services, FP, CCSPT, administration, and non-work-related tasks) during the observation period for each CHW and then dividing the total by 60.
We conducted descriptive statistics, mean estimates for each outcome and associated 95% confidence intervals. We used paired sample t-test to ascertain mean differences between groups before and after CCSPT. We defined statistical significance as a p value < 0.05. For each outcome whose mean significantly changed after CCSPT in these crude analyses, we evaluated the change in time after CCSPT using multivariable repeated measures mixed regression models. This approach accounted for the correlation between observations contributed by each CHW in both the pre- and post-intervention periods [26]. The main independent variable in the models was CCSPT, an indicator variable coded 1 if the observation was during CCSPT implementation and 0 otherwise. Other control variables included age, sex, years of work experience, and facility catchment area population, selected based on existing literature on determinants of CHWs performance or motivations [27, 28]. Given the small number of CHWs observed, we bootstrapped the results based on clustering at the CHW level to estimate parameter standard errors and corresponding confidence intervals [29].
Ethical approval and consent to participate
Ethical approval for the study was provided by National Health Sciences Research Committee (NHSRC) protocol number 19/03/2355 and University of North Carolina (UNC) Institutional Review Board (IRB) protocol number 21094. Verbal informed consent was obtained from all Community Health Workers before start of data collection. The procedure for obtaining verbal informed consent from the participants was ethically approved by both the NHSRC and UNC IRBs. All methods were carried out in accordance with relevant guidelines and regulations.
