A randomised controlled feasibility trial of a BabyWASH household playspace: The CAMPI study

Background Water, sanitation and hygiene (WASH) interventions should support infant growth but trial results are inconsistent. Frequently, interventions do not consider behaviours or transmission pathways specific to age. A household playspace (HPS) is one intervention component which may block faecal-oral transmission. This study was a two-armed, parallel-group, randomised, controlled feasibility trial of a HPS in rural Ethiopia. It aimed to recommend proceeding to a definitive trial. Secondary outcomes included effects on infant health, injury prevention and women’s time. Methods November 2019−January 2020 106 households were identified and assessed for eligibility. Recruited households (N = 100) were randomised (blinded prior to the trial start) to intervention or control (both n = 50). Outcomes included recruitment, attrition, adherence, and acceptability. Data were collected at baseline, two and four weeks. Findings Recruitment met a priori criteria (≥80%). There was no loss to follow-up, and no non-use, meeting adherence criteria (both ≤10%). Further, 48.0% (95% CI 33.7−62.6; n = 24) of households appropriately used and 56.0% (41.3−70.0; n = 28) cleaned the HPS over four weeks, partly meeting adherence criteria (≥50%). For acceptability, 41.0% (31.3−51.3; n = 41) of infants were in the HPS during random visits, failing criteria (≥50%). Further, the proportion of HPS use decreased during some activities, failing criteria (no decrease in use). A modified Barrier Analysis described good acceptability and multiple secondary benefits, including on women’s time burden and infant injury prevention. Interpretation Despite failing some a priori criteria, the trial demonstrated mixed adherence and good acceptability among intervention households. A definitive trial to determine efficacy is warranted if recommended adjustments are made. Funding People In Need; Czech Development Agency. Trial registration RIDIE-ID-5de0b6938afb8.

Indeed, the modest effects on growth in nutrition interventions suggests that a combination of 35 recurrent infections, chronic inflammation, and gut enteropathy limit the effects of nutrition. 6  The evidence on the health and non-health benefits of a HPS or playmat has been previously 57 reviewed. 23 Further, formative data during the participatory design and build of the HPS prototype 58 suggested caregivers liked it and were glad to use it during daily routines. However, there remains a 59 need to assess how long a HPS would be used throughout the day and appropriately maintained and 60 cleaned. Data on infant health outcomes would provide insight into the potential for a HPS to reduce 61 infection from within the home. Moreover, WASH interventions deliver both health and non-health 62 outcomes, all of which contribute to household wellbeing. Thus broader benefits of a HPS, including 63 on women s time and child socioemotional development, also require exploration through a 64 definitive RCT. 65 66

67
The Campylobacter-Associated Malnutrition Playspace Intervention (CAMPI) trial was a randomised, 68 controlled feasibility trial to establish the feasibility of a definitive RCT of a HPS in rural Ethiopia. 69 The HPS design (S1), is described elsewhere, 23 underpinned by previous formative research. 24  study information and participants were given time to make an informed decision. Households were 109 then revisited, eligibility was re-verified, and if households were willing, consent was gained. 110 Households were blinded to their status in the trial until after baseline data collection. Figure 1  111 describes trial enrollment and numbers. 112

Study intervention 116
The trial was conducted in Sidama zone, Januar March 2020. Two field teams managed intervention 117 and control kebeles. After baseline data collection, caregivers from intervention households were 118 called to the kebele health post for a sensitisation da . PIN field team, HEWs and data collectors 119 formally discussed the study rationale, caregiver beliefs around infant faecal-oral transmission and 120 health outcomes, transmission routes and how a HPS might interrupt these to improve infant health. 121 Correct HPS use, maintenance and cleaning was detailed. Caregivers watched and practiced HPS 122 assembly and discussed potential safety issues. Use was discussed in relation to daily routines and 123 activities and caregivers agreed to use it when possible. Households agreed to clean the HPS at least 124 every other day (and always after defecation or urination) with both soap and water. Playspaces were 125 distributed with safety instructions printed in both Sidamo and Amharic with illustrations. HEWs 126 visited intervention households in the following days to ensure correct HPS assembly. The control 127 group received a HPS upon study completion. Households were visited at baseline and at two and four weeks. The primary caregiver present was 132 interviewed, usually the mother. Baseline data included a previously validated survey 25,26 on WASH 133 facilities and use and animal husbandry. Food hygiene, breastfeeding, and diarrhoea incidence were 134 also assessed and again at two and four weeks. Trained data collectors took weight, height and mid-135 upper arm circumference (MUAC) following standard procedure 27 with a digital mother-child smart 136 scale (Ultratec ® ), a foldable infantometer to 5 mm accuracy (seca 210 ® ) and standard MUAC tape to 1 137 mm accuracy, respectively. Seven day diarrhoea prevalence was by caregiver report. Appropriate use/maintenance and cleaning), and Acceptability (random observation of HPS use and 153 change in incidence [proportion] of use from two to four weeks). A modified Barrier Analysis at four 154 weeks provided further insight into acceptability. As these outcomes were the main measures to 155 determine whether to proceed to a definitive trial, a priori threshold criteria were established as 156 follows: 1. Recruitment: the proportion of contacted households participating in the trial would be 157 80%; 2. Attrition: the level at the trial end would be 10%; 3. Adherence: the proportion of non-use 158 of HPS would be 10% at both time points and over the trial; 4. Adherence: the proportion of correct 159 HPS use and cleaning would be 50% at both time points and over the trial; 5. Acceptability: the 160 proportion of infants in the HPS at random check would be 50% at both time points and over the 161 trial, and 6. Acceptability: reported incidence of HPS use during daily activities (as a proportion) 162 would not decrease from two to four weeks. Outcomes would also indicate appropriateness of an RCT 163 and provide recommendations for adjusting the intervention design. 164 165

Statistical analysis 166
Data were managed in Excel and analysed in SPSS (v26, IBM). Descriptive statistics summarised 167 survey data and health outcomes. Trial outcomes are displayed with 95% confidence intervals (CI). 168 The adherence outcome included Appropriate use and Appropriate cleaning , created as composite 169 binary outcome variables (described in table footnotes) and described across study time points. 170 Adherence as HPS non-use was described as reported non-use after baseline. Acceptability as 171 Infant in playspace upon arrival was calculated for both visits. Acceptability as change in HPS use 172 was calculated from reported HPS use during reported daily activities over two and four weeks and 173 the difference in proportions. A Generalised Estimating Equaton (GEE) was used as a semi-174 parametric model, using a robust variance estimator and an unstructured working correlation matrix. 175 A binary logistic GEE estimated factors associated with Appropriate use and Appropriate cleaning 176 at two and four weeks. Models were initially run separately: however the merged composite variable 177 of Appropriate use and cleaning showed no difference in parameter estimates between models and is 178 presented. Pre-specified variables included infant sex and age; maternal age; maternal education; 179 number in household; number of children; household owns soap; safe water storage; animal 180 husbandry practices; water availability, and mother collects water. Results are expressed as populated 181 averaged odds ratios (ORs) with 95% CI. information (S1). Derivation of themes was data-driven, where codes resulted from the analysed data 186 as they related to each determinant (See S1 and S6). Coded themes are discussed as either barriers or 187 enablers to the implantation of, and improving outcomes during, a definitive trial. For secondary 188 health outcomes, anthropometric z-scores were calculated (WHO Anthro v3·2·2) and categorised into 189 stunting and wasting using standard cut-off values. 28 Samples positive for presumptive 190 Campylobacter spp., colonies were counted using OpenCFU. Change in diarrhoeal and 191 Campylobacter prevalence between study groups was estimated using a GEE intercept-only model 192 with OR and 95% CI.

Trial outcomes 221
For ease of assessment, study outcomes are described together in Table 2

Recruitment and attrition 234
Rates for recruitment and attrition are shown in Table 2. One hundred households were recruited from 235 four kebeles. To achieve this, 106 households were assessed for eligibility; four households were then 236 excluded for not meeting infant age criteria at the study start and a further two did not consent to 237 participate (Figure 1). Thus a recruitment rate of 94·3% (95% CI 88·1 97·9) met a priori criteria of 238 80%. All households completed the trial assessments at four weeks and there was no loss to follow-239 up (0·0%; 95% CI 0·0 3·6), meeting criteria for attrition ( 10% at trial end). 240 241 Adherence 242 Adherence was first described as the proportion of HPS non-use at both time points and over the study 243 period (Table 2). No households reported not using the HPS at either time point or over the study 244 duration (0.0%, 95% CI 0·0 0·71), meeting a priori criteria 10%. Second, adherence was described 245 through Appropriate use and Appropriate cleaning and combined, across the study time points and 246 throughout the trial (Table 3). Appropriate use included maintenance, as described in the table  247 footnotes alongside variable components (also in S3). When considering behaviours and time points 248 separately, Appropriate use and cleaning were consistently above the a priori threshold of 50%. 249 However when assessing throughout, findings are mixed. Appropriate use did not meet the threshold 250 (48·0%) whilst cleaning did (56·0%) and only 26·0% of households appropriately used and cleaned 251 the HPS throughout the trial. Variables associated with adherence outcomes across the two time 252 points were assessed using a binary logistic GEE model (Table 4). Results display the 95% CI for the 253 effect size and odds ratio. The only variable to significantly predict Appropriate use or cleaning was 254 Mother collects water alone , where an inverse relationship showed a reduced odds of 72·0% (0·28; 255 95% CI 0·12 0·66). 256   The first measure noted if the infant was in the HPS during a random visit (Table 2). This increased 276 from 32·0% (95% CI 19·5 46·7, n=16) at two weeks to 50% (95% CI 35·5 64·5, n=25) at four 277 weeks, meeting a priori criteria of 50% at this point: however throughout the trial did not reach the 278 threshold (41·0%, 95% CI 31·3 51·3; n=41). Second, change in incidence (as a proportion) of HPS 279 during daily activities was assessed. Primary caregivers were asked open-ended questions to record 280 their activities during the past 24 hours, and if they did or did not use the HPS. Results are shown in 281

304
Acceptability was further assessed through a semi-structured questionnaire as a modified Barrier 305 Analysis. This assessed 12 categories of behavioural determinants, exploring all factors which would 306 act as barriers or enablers during a definitive trial (S1). Full results are in S6. The first seven 307 determinants quantitatively assess beliefs and behaviours relating to infant health and HPS use. The materials, e.g. brushes (24·0%, n=12). Importantly, having no older children to watch the infant was a 325 barrier (32·0%, n=16) and relates to the burden of workload on women. A lack of toys was also a 326 barrier (32·0%, n=16). Whilst the design appeared largely acceptable, some difficulties included 327 fitting the rope connecting walls (38·0%, n=19; see S3).

341
No adverse events were observed from HPS use in the intervention group. No household reported any 342 safety concerns associated with use, aside from one household who mentioned the plastic mattress 343 became hot under the sun. HPS use did not increase the risk of any adverse infant health outcome, 344 where the direction of effect does not show an increased risk for the intervention group. 345 346

347
The CAMPI trial is the first randomised, controlled feasibility trial of a HPS in rural, subsistence 348 agriculture households in Ethiopia. Though trial outcomes did not fully reach a priori criteria, a 349 definitive RCT for efficacy is feasible. Results echo two similar studies. In the SHINE trial in 350 Zimbabwe, an imported plastic HPS and locally-sourced plastic playmat were included in a WASH 351 intervention to improve growth and anaemia. Whilst fidelity of delivery was high, 10 the WASH 352 intervention did not prevent infection. 30 However, the analysis did not estimate a magnitude of effect 353 from the HPS specifically. In Zambia, a community-built HPS was assessed alongside a plastic model 354 for acceptability and feasibility. 31 Reported use was similar between the two types (ranging from 10 355 minutes to three hours), family and community reactions suggested acceptability was high and 356 caregiver reports suggested the community built space prevented infant ingestion of soil and animal 357 faeces. Thus growing evidence supports wide acceptability and feasibility across different contexts 358 and further rigorous assessment of efficacy is merited. 359 360 Addressing barriers to appropriate use and cleaning of the HPS would improve these outcomes. Data 361 here described a broadly consistent pattern over the four weeks, albeit with a small decline (Tables 2  362 and 3). The modified Barrier Analysis offered reasons for diminishing use and drops in compliance, 363 including the expense of soap and other cleaning materials. Providing these alongside the HPS would 364 be a key consideration for any future RCT to ensure good hygiene. Similarly, contextual WASH 365 factors, such as water quality, availability, and unsafe storage (76·0%, n=38 in the intervention group; 366 Table 1) must be considered which may result in increased bacterial transmission. Similarly, the team 367 decided not to provide toys during the trial given the potential to become vectors for indirect faecal-368 oral transmission. 18,32 However, this was a frequently cited barrier for mothers whose infants became 369 choice, reducing anxiety, and even freeing up time to spend with her infant. However any negative 387 long-term impacts will need to balance these. This includes a lack of infant supervision, and the risk 388 of reinforcing women s roles as sole caregivers alongside a continuing responsibilit for other 389 domestic duties. This is reinforced by the GEE model (Table 4) where when the mother bore the duty 390 of collecting water alone, the HPS was less likely to be used or cleaned properly. In many low-income 391 countries, women s triple work burden in the productive, reproductive and social domains impedes 392 their well-being and may reduce engagement in childcare 33 a pattern often inherited by older female 393 siblings. This highlights a trade-off in encouraging more active parenting alongside existing home 394 duties, and any intervention must ensure it does not further encumber women. 395

396
The CAMPI trial was not powered to detect any differences in health outcomes between groups and 397 results should be interpreted accordingly. However secondary infant health outcomes indicated the 398 potential efficacy of a HPS and appropriateness of these outcomes for a future RCT. Diarrhoeal 399 prevalence from baseline reduced among the intervention group whilst presumptive Campylobacter 400 did not (Table 6). Beyond the lack of adequate power, substantial methodological limitations may 401 profoundly affect validity: these include the reliability of caregiver-reported diarrhoea and a 402 desirability bias within intervention households. No further GEE analysis was performed to explore 403 associated variables. However, aside from a potential lack of effect of the HPS on Campylobacter 404 prevalence, other pathways not interrupted by the HPS likely contributed to pathogen transmission. 405 This includes incorrectly (re-)heated foods 34 ; data on this indicated unsafe practices were common 406 (S8) where across households only 28 safely prepared all meals at both time points (data not shown). 407 All infants were given liquids other than breastmilk, including water, possibly contaminated through 408 unsafe storage or other pathways. Campylobacter from domestic free-range poultry appears to present 409 an infection risk to infants 5,26 and here poultry frequently shared living and sleeping areas (Table 1).

Progression to a definitive RCT 416
To improve playspace adherence and acceptance, a future definitive RCT should focus on directly 417 addressing the barriers whilst promoting the enabling factors as identified in this feasibility trial. 418 Whilst further behavioural modules and developing caregiver knowledge might have improved 419 outcomes, it is not always practical. During the sensitisation day the HPS was introduced in a 420 scalable manner to reduce work burden among households and HEWs who are already overworked. 421 Rather, to achieve behavioural change it is pragmatic to directly address barriers and promote 422 enabling factors. Knowledge alone is unlikely to prevent infant faecal-oral transmission without a 423 material element which breaks contact, and an enabling technology may drive changes in behaviour 424 but still requires addressing factors which support or obstruct change. Factors included in the 425 composite variable Appropriate use responsible for a decline include another child sharing the HPS 426 (S3). Given the potential to introduce contamination, this might be addressed by a visiting HEW as a 427 risk factor. Similarly, Appropriate cleaning declined from every day/every other day to twice a week. 428 The direction of effect and significance in the GEE model (Table 4)  number of data collectors for the sample. However, it is important to note that daily data collection 438 was intense and required serious team dedication. A larger trial would likely experience higher drop 439 out without equivalent input: a 95% CI estimate would be between 96 100% in a power calculation 440 and 95 100% if repeated maintaining the same effort and ratio of stud personnel. Over a longer time 441 period, this is likely unsustainable. Future sample size calculations must consider these number 442 requirements for study personnel. Furthermore, as recipients of previous WASH interventions, the 443 intervention group likely adopted the new intervention modality earlier than might be seen in other 444 contexts, holding implications for external validity. Good uptake may also be seen in other contexts 445 where NGOs have a known presence and have provided multiple interventions for many years, but 446 this does limit the generalisability of findings to other contexts. Lastly it is important to note the 447 extensive HPS design process and the underlying formative work. Good contextual understanding is 448 critical for intervention success, which must be culturally acceptable, locally integrated and must 449 consider contextual baseline demographic and WASH characteristics and health status which vary 450 significantly. 451

453
The CAMPI trial evaluated feasibility of a BabyWASH HPS and recommendations to progress to a 454 future RCT in a rural, subsistence agriculture setting in Ethiopia. Not all a priori criteria were met. 455 Overall the HPS showed mixed engagement and adherence, good acceptability and many reported 456     Activities .

Barrier Analysis method in the CAMPI feasibility trial
Given that there were no non-doers in the CAMPI feasibilit trial, intervention households were not analysed according to the standard Doer/Non-doer methodology as described above. Rather, the methodology was modified slightly to explore behaviours among all participants, without categorising them as Doers/Non-doers. The survey used in the modified BA is shown below. Intervention households were interviewed by the data collection teams, trained in the BA method, alongside a Health Extension Worker at week four of the feasibility trial. Following data collection, responses were entered into Excel. Questions 2 9 were entered into categories according to determinants and then categorised into themes which arose from the data. Those themes were summed and described in the results. Quantitative responses (questions 1, 10 16 in the table below) were summed in each answer category and also subsequently described in the result. As such, the determinants allowed for the description of barriers/enabling factors among all households which would improve adherence to a greater degree given that all households used the playspace during the trial. What are the advantages of using the play space? 3 What are the disadvantages of using the play space? 4 What makes it easy for you to use the play space? 5 What makes it difficult for you to use the play space? 6 What makes it easy for you to keep the play space clean? 7 What makes it difficult for you to keep the play space clean? 8 Who are the people who approve of you using the play space for your child? 9 Who are the people who disapprove of you using the play space for your child?   Feeding of fresh or reheated foods prepared as recommended if: Meals were prepared less than 2 hours before eating (not reheated); Meals were prepared less than 2 hours before eating which were reheated to almost boiling ; Meals were prepared more than 2 hours before eating but reheated to almost boiling .