Bulletin of the World Health Organization

Observations of infection prevention and control practices in primary health care, Kenya

Guadalupe Bedoya a, Amy Dolinger a, Khama Rogo b, Njeri Mwaura b, Francis Wafula b, Jorge Coarasa c, Ana Goicoechea d & Jishnu Das a

a. Development Economics Research Group, The World Bank, 1818 H St NW, Washington DC, 20433, United States of America (USA).
b. Health, Nutrition and Population Global Practice, The World Bank, Nairobi, Kenya.
c. Health, Nutrition and Population Global Practice, The World Bank, New Delhi, India.
d. Trade and Competitiveness Global Practice, The World Bank, Washington DC, USA.

Correspondence to Jishnu Das (email: jdas1@worldbank.org).

(Submitted: 31 May 2016 – Revised version received: 20 December 2016 – Accepted: 21 December 2016 – Published online: 09 March 2017.)

Bulletin of the World Health Organization 2017;95:503-516. doi: http://dx.doi.org/10.2471/BLT.16.179499

Introduction

The prevention and control of infections is critical for a well-functioning health system. However, worldwide an estimated 21 million cases of hepatitis B virus infection and 200 000 cases of human immunodeficiency virus (HIV) infection result from unsafe injection practices each year.1 In the United States of America, an estimated 40 000 to 80 000 deaths are due to nosocomial infections annually, which may cost as much as 4.5 billion United States dollars (US$).2 Moreover, the rapid spread of multidrug-resistant organisms and outbreaks of Ebola virus disease, yellow fever and Zika virus infections has further increased the human and financial cost. Fortunately, proven and cost-effective, infection prevention and control practices can reduce the risk.36

However, ensuring compliance with these practices depends on understanding the extent of the problem and there has been little research on infection prevention and control practices in low- and middle-income countries, particularly in primary health care.710 Previous studies have several limitations. First, many involved small samples – one review found that only 10 of 41 studies on hand hygiene interventions were conducted in more than one hospital.10,11 Second, they were often based on self-reported data from health-care providers, which tend to overestimate compliance.12 Third, they frequently focused on single domains, such as injection safety, rather than on the range of possible exposures encountered by patients during outpatient visits.1,1317 Consequently, given that infections are transmitted by multiple pathways, these data are of limited use for modelling the spread of different pathogens.10,18

To address these limitations, we designed a novel observational tool. We based the tool on existing World Health Organization (WHO) tools that can track patients throughout the course of an outpatient visit, which may include examinations, laboratory tests and injections. Our tool can be used to assess compliance with infection prevention and control practices throughout primary health care and can help identify associations between compliance and individual characteristics of health-care facilities and workers.

We used the observational tool to investigate compliance with infection prevention and control practices in a pilot sample of 23 health-care facilities in Nairobi, Kenya and, subsequently, in 935 facilities in three Kenyan counties.

Methods

The design of the observational tool took into account the possibility that an outpatient may experience a violation of infection prevention and control practices in several different locations: the consulting room, the laboratory and the injection room. In this study, trained assessors spent three consecutive hours in each of 958 health-care facilities. The observational tool was used to record all interactions between patients and at least one health-care worker at each location. In collaboration with the Kenyan Ministry of Health and local experts, we identified three key procedures for observation: (i) physical examination; (ii) injection; and (iii) blood sampling. Five infection prevention and control domains were observed across these procedures: (i) hand hygiene; (ii) use of protective gloves; (iii) injection and blood sampling safety; (iv) disinfection of reusable equipment; and (v) waste segregation. These domains have been identified as critical for outpatient safety by WHO,19 the United States Centers for Disease Control and Prevention,20 the WHO Regional Committee for Africa21 and the Kenyan Ministry of Health.22 In accordance with the broad consensus that safety of care is a characteristic of the system and not just of individual health-care providers,23 for each of these domains, we assessed: (i) compliance with infection prevention and control practices by health-care workers; (ii) knowledge of these practices among health-care workers; and (iii) the availability of the equipment and supplies needed for implementing these practices.

Assessing compliance

The assessment of compliance with infection prevention and control practices was based on indications and corresponding actions. An indication refers to a situation in which an infection prevention and control practice must be undertaken to prevent the risk of a pathogen being transmitted from one surface to another. Actions occur in response to indications, such that each indication has a corresponding action. Compliance means that the correct action has been taken. For example, for the domain of hand hygiene (Table 1), the indication “Before touching the patient” indicates the possibility that physical contact could lead to microbial transmission. The correct action corresponding to this indication is: “Health-care worker washed his or her hands with soap or used an alcohol-based hand rub.”2426 We determined the proportion of indications that were accompanied by the corresponding action for all health-care workers across each of the 20 indications listed in Table 1 for five infection prevention and control domains. For example, for the domain of hand hygiene, compliance with seven well-known indications was assessed. Table 1 also lists the equipment and supplies essential for carrying out each action and Table 2 describes how health-care workers’ knowledge of the actions required in each domain was assessed. The novelty of our approach is that we used a single observational tool to collect comprehensive data on indications and their corresponding actions across a large number of infection prevention and control practices in an outpatient setting in a low-income country.20,27,28

A strict survey protocol ensured that study assessors did not interact with health-care workers or patients during observation. Data quality was assured using several methods: (i) the assessors’ skills were evaluated using tests and videos; (ii) data forms were reviewed daily; (iii) inter-rater reliability was evaluated on a sample of observations; and (iv) data were double-entered to ensure an error rate below 1%. Data collected by the observational, patient-tracking tool were recorded on paper and tablet computers were used in assessing the availability of supplies and health-care workers’ knowledge.

Health-care facilities

The observational tool was piloted in 605 patients at 23 health-care facilities in Nairobi, which were selected to represent the wide range of facilities that provide primary health care in Kenya. They included public, private not-for-profit and private commercial facilities of differing levels of complexity: level-2 facilities included basic dispensaries and clinics, whereas level-5 facilities included county referral hospitals offering both basic and specialized services.

The main study took place in Kakamega, Meru and Kilifi counties in different regions of Kenya. Consent was obtained from 1035 of the 1115 (93%) health-care facilities identified and 935 took part in the study: 94 had no patients on the survey day and 6 did not provide consent on that day. In addition, 99% (14 443/14 531) of patients and 100% (1680/1680) of health-care workers approached also consented to being observed. The characteristics of participating facilities, health-care workers and patients are listed in Table 3. Although the distribution of facilities by level of complexity was similar to that at the national and county level, the proportion of private facilities was higher, possibly because these facilities were under-represented in administrative data.

Overall, we observed 21 791 procedures in the main study (i.e. physical examination, injection or blood sampling) performed by 1680 health-care workers in 14 328 patients and we registered 106 464 indications for infection prevention and control practices. In higher-level facilities (i.e. levels 3, 4 and 5) with two or more health-care workers, we observed each worker for at least 1 hour. The mean age of the health-care workers was 37.4 years, they worked a mean of 5.43 days a week in their facilities and 51% were male (Table 3). Of the patients, 40% (5664/14 328) were male and 52% (7441/14 328) were younger than 18 years. Because public facilities see more patients, 70% (9976/14 328) were observed at public facilities even though only 39% (369/935) of facilities were public.

The study was approved by the Ethics and Scientific Review Board at the African Medical and Research Foundation (Approval no. AMREF-ESRC P94/2013), the Kenyan Ministry of Health and authorities at participating facilities.

Statistical analysis

We calculated the compliance by dividing the number of correct actions performed by the total number of indications and report the observed compliance with infection prevention and control practices for each domain (i.e. aggregated across indications) and for all domains combined. This implies that the implicit weight given to each individual action is the frequency with which it was observed. We adopted this approach because the relative risk associated with different infection prevention and control practices in primary health care has not been established. To assess whether compliance was associated with the individual characteristics of the health-care worker or facility, we proceeded in a stepwise fashion. First, we used multiple linear regression to assess the association between the availability of supplies and compliance. Then, we used a similar analysis to assess the association between other characteristics and compliance in the subset of observations for which supplies were available since the compliance would necessarily be zero if essential supplies were unavailable. In particular, it can be shown that, when supplies are necessary but not sufficient for compliance, the lower bound for the association between the availability of supplies and compliance is the mean compliance in the sample. Consequently, the regression coefficient for the association between the availability of supplies and compliance will lie between the mean compliance in the sample and 1. In deriving standard errors, we adjusted for the effect of clustering at the level of the facility and of the health-care worker. Data were analysed using Stata version 13.0 (StataCorp. LP, College Station, United States of America). Further details of the statistical methods are available from the corresponding author.

Results

In the pilot study, we found that: (i) the level of consent was high, with 98% (605/617) of patients and 98% (50/51) of health-care workers consenting; (ii) the use of identification tags to track patients across different units in a facility was effective; and (iii) inter-rater agreement on assessments was high (kappa: 0.72; 95% confidence interval, CI: 0.68 to 0.76), with 93% of comparisons in agreement. The mean overall compliance across the five infection prevention and control domains was 0.382 (95% CI: 0.366 to 0.399) and the mean number of safety violations per patient was 3.7 (95% CI: 3.6 to 3.8). The mean compliance was highest for the injections and blood samples domain, at 0.988 (95% CI: 0.980 to 0.996), and lowest for hand hygiene, at 0.028 (95% CI: 0.020 to 0.037).

In the main study, the mean number of infection prevention and control indications per patient was 7.5 (95% CI: 7.4 to 7.6) and the mean number of safety violations per patient was 5.1 (95% CI: 5.1 to 5.2) for each outpatient visit. The mean overall compliance was 0.318 (95% CI: 0.315 to 0.321) for the 106 464 indications observed. The number of indications and safety violations increased with the number of procedures but compliance varied according to the specific procedures performed (Fig. 1). This observation is consistent with the substantial variation in compliance across domains (Fig. 2). For example, the mean compliance in the injections and blood samples domain was 0.871 (95% CI: 0.866 to 0.876) compared with 0.023 (95% CI: 0.021 to 0.024) in the hand hygiene domain.

Fig. 1. Infection prevention and control indications and safety violations, infection prevention and control study, Kenya, 2015
Fig. 1. Infection prevention and control indications and safety violations, infection prevention and control study, Kenya, 2015
CI: confidence interval.
Notes: An indication refers to a situation in which an infection prevention and control practice must be undertaken to prevent the risk of a pathogen being transmitted from one surface to another (Table 1). A safety violation occurred when the required action was not taken. The percentages in parenthesis give the proportion of patients who underwent the procedure or combination of procedures.
Fig. 2. Compliance with infection prevention and control practices, by infection prevention and control domain, Kenya, 2015
Fig. 2. Compliance with infection prevention and control practices, by infection prevention and control domain, Kenya, 2015
CI: confidence interval.
a The compliance is the proportion of indications for an infection prevention and control practice for which the corresponding action was taken.
Notes: An indication refers to a situation in which an infection prevention and control practice must be undertaken to prevent the risk of a pathogen being transmitted from one surface to another (Table 1). The data in the figure relate to all 106 464 indications observed.

We found small, weak associations between compliance and most characteristics of health-care workers and facilities. The proportion of indications for which the health-care worker had the requisite knowledge and the proportion for which the requisite supplies were available were considerably higher than the rate of compliance for those indications in most domains (Table 4). In the hand hygiene domain, the mean compliance was 0.042 when the health-care worker had the requisite knowledge and the requisite supplies compared with an overall mean of 0.024. The regression coefficient for the association between compliance and the availability of supplies determined using the ordinary least squares method is 0.368 (95% CI: 0.352 to 0.385; Table 5), which was very close to the observed overall mean compliance of 0.318. However, after domain fixed effects were taken into account, the regression coefficient was 0.162 (95% CI: 0.145 to 0.179), which suggests that supplies were necessary but not sufficient for compliance.

For the 68 034 observed indications for which supplies were available, the average compliance was 0.390 (95% CI: 0.386 to 0.393). Table 5 lists regression coefficients for the association between compliance and other health-care worker and facility characteristics in the subset of 64 494 observations for which supplies were available and data on health-care workers’ knowledge were recorded. Regression coefficients were derived after separately taking into account domain fixed effects and adding health-care worker fixed effects, which enabled us to assess whether differences in the supplies available or in knowledge between individual health-care workers were associated with compliance. In this subset of observations, the estimated regression coefficients for compliance, after domain fixed effects were taken into account, were most strongly associated with the domain – the inclusion of indicator variables for the domains increased the adjusted R2 (which indicates how close data are to the fitted regression line) for the regression from 0.154 to 0.650. In contrast, compliance was weakly associated with the type of facility (i.e. public or private or level of specialization), the health-care worker’s educational level, age and sex, the availability of Kenyan infection prevention and control guidelines (only 5% of facilities had a copy) and whether the health-care worker had undergone training on infection prevention and control within the last year. Also in the subset of observations for which supplies were available, the regression coefficient for the association between compliance and health-care workers’ knowledge was 0.035 (95% CI: 0.021 to 0.050) after domain fixed effects were taken into account and 0.026 (95% CI: 0.011 to 0.041) once health-care worker fixed effects (e.g. an individual worker’s level of motivation) were taken into account.

Table 6 shows the association between compliance and the availability of supplies separately for each infection prevention and control domain: the regression coefficient for compliance ranged from 0.006 (95% CI: −0.059 to 0.071) for the reusable equipment domain to 0.848 (95% CI: 0.801 to 0.894) for the waste segregation of needles and syringes domain. Regression coefficients for the association between compliance and health-care workers’ knowledge when supplies were available (Table 7) were smaller and few were statistically significant. The largest coefficients were 0.184 (95% CI: 0.066 to 0.302) for the waste segregation of needles and syringes domain and 0.151 (95% CI: 0.083 to 0.219) for the reusable equipment domain.

In domains, such as the hand hygiene domain, where the mean compliance and estimated regression coefficients for the association between compliance and the availability of supplies and knowledge were all small, it is unlikely that a lack of supplies or knowledge was the only constraint. In contrast, in domains such as the waste segregation excluding needles and syringes domain, where the regression coefficient for the association between compliance and the availability of supplies was higher and the observed compliance was small, it is likely that the availability of supplies was a more important constraining factor.

Discussion

Our observational, patient-tracking tool was able to assess compliance with infection prevention and control practices across a range of health-care facilities and infection prevention and control domains. We found there were, on average, 13 opportunities for infection transmission during an outpatient visit if the patient was examined, had an injection and underwent blood sampling. Compliance varied widely across indications: it was almost complete for single-use needles and syringes but very low for hand hygiene – when practiced, hand washing lasted an average of 16 s compared with the recommended 30 to 60 s. Data collected using the tool can provide key information for epidemiological modelling of disease outbreaks because different safety violations may be associated with different risks depending on the disease transmission mechanism.29,30

We found only weak associations between compliance and the availability of supplies, health-care workers’ knowledge, training in infection prevention and control and the availability of guidelines. These findings are consistent with the widely discussed concept that patient safety is driven more by behavioural norms than by technical knowledge, training or the availability of supplies.12,31 Consequently, compliance depends on engendering these norms, which has been achieved in Kenya for injections and blood sampling but not for hand hygiene. There has been substantial decline in unsafe injection practices worldwide,32 which could be extended to other practices. In Australia, for example, altering behavioural norms in hospitals has substantially improved hand hygiene.33

Our observational tool for assessing infection prevention and control practices across multiple domains has several limitations. First, linking practices to health outcomes requires data on the types of pathogens present at observation sites – this would ultimately enable researchers to apply an appropriate weighting to compliance with specific infection prevention and control indications. However, there are no literature reports on the relative risks of different practices, even in high-income countries. Moreover, although WHO suggests that it is reasonable to focus on key domains that are consistently linked with nosocomial infections (e.g. hand hygiene is considered to be the single most effective infection control measure),1 the evidence from outpatient settings is sparse. Second, our tool was based on direct observations, which could have been subject to the Hawthorne effect, whereby health-care providers changed their behaviour when observed. However, previous studies of this effect suggest that observation-induced behaviour decreases with the number of interactions observed.34 We examined whether early observations differed from later observations and, although we found a small, negative association, it disappeared once we controlled for the infection prevention and control domain and health-care worker fixed-effects. The absence of a Hawthorne effect is encouraging because some researchers consider direct observation to be the gold standard for measuring compliance with infection prevention and control practices since it makes it possible to record both indications and their corresponding actions.20,25 Finally, our data may be incomplete because they relate only to the day of observation. For example, waste disposal (not waste segregation) may take place on only one day of the week or month. Although we were unable to comprehensively analyse waste disposal, there was evidence of important gaps in waste management. For instance, only 11.1% of facilities had a standard operating procedure for waste management, only 26.1% had an on-site incinerator or a contract with a company for incineration and only 27.8% had a waste holding area.

In conclusion, our observational, patient-tracking tool provided an effective way of assessing compliance with infection prevention and control practices across multiple domains in primary health care. It could be used to rapidly assess the current status of these practices and to monitor improvement efforts. We found that compliance with infection prevention and control practices was low overall but varied substantially across domains. The variations were only weakly associated with the characteristics of the facility and the health-care worker, such as the health-care worker’s knowledge and the availability of supplies, which suggests that improvements will require a broader focus on behavioural change.


Acknowledgements

We thank: Rachael Kamau, Charles Kandie, Mary Wangai and John Kabanya at the Kenyan Ministry of Health; Scott Lee at Brigham and Women's Hospital; Linus Ndgewa, IPC Coordinator at CDC Kenya; Pamela Njuguna and Bernard Olayo at the World Bank Group; Brenda Barasa (Kakamega), Omar Swabah and Anisa Omar (Kilifi) and William M Muraah and James Gitonga (Meru) of the county health teams; Sherlene Chatterji, Benjamin Daniels, Annie Kong, Shubhra Mittal and Garima Sharma, the IPC Unit at Kenyatta National Hospital, staff of subcounty health management teams, IPSOS staff and Innovations for Poverty Action (IPA) staff.

Funding:

This paper used data from the Kenya Patient Safety Impact Evaluation (KePSIE) project – a joint undertaking between the World Bank Group and the Kenyan Ministry of Health. Funding was provided by the World Bank Group through the Strategic Impact Evaluation Fund, the Impact Evaluation to Development Impact fund, the Trade and Competitiveness Impact Program, the Knowledge for Change Program and the Primary Health Care Performance Initiative.

Competing interests:

None declared.

References