The Harti RCT is grounded in Kaupapa Māori Theory. This includes recognising the existence and validity of Māori knowledge, language and culture (12). Kaupapa Māori approaches to research have been described as “by Māori, for Māori, with Māori”, that Māori self-definitions and self-valuations are affirmed, as ensuring that tikanga Māori (practices and customs) and processes are followed throughout the research, Māori benefit from the study, and that colonial constructions and definitions of Māori are critiqued (13).

Kaupapa Māori Theory within the Harti programme recognises the need to critique conventional approaches, enact systems change, direct structural transformation; and address unequal power relationships to improve outcomes for Māori (12). Our study is underpinned by the He Pikinga Waiora (HPW) Implementation Framework—a theoretical and practical guide for developing and implementing interventions that has Māori self-determination at its core and includes co-design methods (14, 15). Co-design is an important best practice principle for Kaupapa Māori research (16). In part this is due to the iterative approach of co-design that allows for conceptual re-developments and refinement, based on the social-cultural needs of partnership groups (11). In line with the broader methodological Kaupapa Māori approach, the Harti process takes an ecological system-focused approach (17) to research questions, whereby our interest is understanding the complex environment, and supporting health outcome improvements at both the family (whānau) level as well as at the systems level, rather than a focus purely on individual patients’.

Holistic, whānau-centred care approaches have long been integral to Māori conceptualisations of health and wellbeing (18). The NZ Ministry of Health's Māori Health strategy He Korowai Oranga (2002) and Whakamaua: Māori Health Action Plan aim for a health system that will work in a way that acknowledges Māori health aspirations and the central role of whānau as a principal source of strength, support, security, and identity (19, 20), Family or whānau-centred practice is also being increasingly recognised as a key goal for modern healthcare. Whānau-centred practice is characterised by a “partnership between parents and service providers, a focus on the family's role in decision-making, and recognition that parents are the experts on their child” (p.357) (21). While the concepts of family-centred practice are often described within NZ's education, health, and social welfare sector policy documents (22), a disconnect has been identified between these descriptions of best practice and what in fact occurs (23).

The study uses a Kaupapa Māori methodology with qualitative and quantitative methods. Qualitative methods include in-depth interviews with whānau and are published elsewhere (24). Quantitative methodology was a randomised controlled observer-blinded parallel groups single centre superiority trial. It is designed to measure the impact of the Harti tool, compared with usual hospital care, on child and whānau wellbeing outcomes, as specified in the primary and secondary endpoints. Tamariki and their whānau will be randomised to the intervention and control groups at a ratio of 1:1. The intervention group will receive usual care plus the Harti tool, and the control group will receive usual hospital care only, as per the trial schematic shown in Figure 1.

A co-design process was initiated as a response to issues of unmet health need and inequitable and unacceptable hospital readmission rates for tamariki Māori. The process was led by staff at Te Puna Oranga (the Māori Health Unit) at Te Whatu Ora Waikato, who worked with nurses, Kaitiaki (cultural support workers) and doctors who had worked or were working in child health at Waikato Hospital (11). The team identified that hospital care focused almost exclusively on single illnesses and that the model of care did not able the root causes of disease to be address, nor the broader opportunities to improve wellbeing for tamariki and their whānau to be realised. The team concluded that a holistic and whānau ora approach was needed for hospitalised tamariki and their whānau. Whānau ora describes the healthy families goal within New Zealand's Māori Health Strategy, He Korowai Oranga (25), and is also a government work programme. For this project we use the term whānau ora to describe an approach which places whānau at the centre of service delivery, requiring the integration of health, education, and social services around needs of whānau.

There is increasing recognition worldwide of the importance of wellbeing care needs (such as addressing poverty, education, housing, social and cultural cohesion) in policies and programmes aimed at improving health outcomes (26–30). There is also a developing understanding that the health care setting provides a unique opportunity to address both health and broader wellbeing care needs by improving the coordination and quality of health care; and more effectively supporting priority populations.

Over a 1 year period, the iterative co-design process resulted in the development of a tool to identity health and wellbeing needs, and to support opportunities to meet needs and enable the provision of holistic and preventative care particularly for hospitalised tamariki Māori and their whānau. The tool, developed for all tamariki hospitalised at Waikato Hospital, was named the Harti Hauora Tamariki Tool by the then General Manager of Te Puna Oranga, Māori Health. First iterations of the tool had sections for completion by both nursing and medical staff. However, the medical staff were not consistent in completion of their sections. The final iteration was completely a nursing document. The Harti tool was printed onto a card and included in the front of medical notes files. A staff training folder was developed, and training sessions were held with ward staff and a training folder was kept in each ward. A local Māori artist helped develop a whānau Harti information booklet with information about available services—which was given to all whānau along with a variety of resources for tamariki.

The Harti Tool was used in the paediatric medical wards at Waikato Hospital from 2014 to September 2017. The paediatric medical charge nurse reported that the majority (over 90%), of patients and their whānau were assessed using the tool during in that period. The Harti tool was used for the majority of tamariki and their whānau during this time.

Recruitment and baseline data collection will be the responsibility of a RA who will be present, when possible, on the wards seven days per week. At the start of each day, a RA will generate a list of patients currently on the paediatric ward, screen them for eligibility, and check whether they have already been consented. RAs will check with the ward charge nurse to see if it is appropriate to approach the whānau of eligible patients. A record will be kept of all patients screened for eligibility to be approached and participate in the trial. Reasons for non-eligibility and/or refusal to participate will be recorded in a tracking sheet. Each RA will assure whānau that they have no obligation to provide a reason for non-participation in the trial. At the time of obtaining informed consent a baseline enrolment ascertainment form will be completed. The ascertainment form will record socio-demographic information (age, ethnicity and gender of tamariki patient and caregiver), method of transport to hospital, household size and structure, housing type and number of bedrooms, and level of material deprivation or NZiDep (33). The ascertainment form will be entered online using Qualtrics™ survey software.

A questionnaire on patient and whānau satisfaction with hospital experiences will be administered prior to discharge, or within 2 weeks of discharge. We will use the Mārama Real-Time Feedback tool, which is used extensively throughout health services in NZ. It was developed in 2016 by CBG Health Research as part of a one-year pilot funded by the Health and Disability Commissioner (HDC) (34).

Randomisation will proceed according to a strict simple (non-stratified) protocol, following consenting, using the Qualtrics™ Survey software which uses the Mersenne Twister algorithm to determine allocation to intervention or control group. The RA will not reveal allocation to the participant, nor will anyone else. The only other person who will have access to group allocation data is the Study Manager.

The RA will continue with the Harti tool or inform usual care participants that they will return to go over the final short satisfaction with hospital experiences Mārama survey before they are discharged.

Because the intervention is whānau/family focussed, study contamination will occur if members of the same household are allocated to different intervention groups. Thus, for any child where another sibling or tamariki/child living in the same household has previously been consented and randomised in the trial, the subsequent child will be allocated to the same intervention group as their household member.

We aim to recruit a total of 1,000 participants, of whom we expect approximately 40% will be Māori (based on current figures and expected acceptability of the research protocol). This sample has been calculated to ensure that there is sufficient statistical power to detect a 7% absolute reduction in the 30-day readmission risk in the intervention group for all children and a 12% absolute reduction in the 30-day acute readmission risk for Māori. Sample size calculations are based on a two-tailed 0.05 level of significance with 80% power. We assume that the acceptability of participation will be similar for Māori and non-Māori.

We will measure differences between the intervention and control groups in the primary endpoint (readmission risk) controlling for age, sex, residing in Hamilton, socioeconomic deprivation (NZiDep), and RA responsible for conducting the intervention. Analyses of the primary endpoint will be stratified by ethnicity.

The adult/s accompanying the child and the child will be approached by the RA who will engage using the process of whānaungatanga and values of whakamana and manaakitanga. The RA will introduce herself/himself, and explain the study in broad terms, saying what participation may involve, including matching the child's National Health Index (NHI) number against health outcome databases, plus giving the opportunity to ask questions and give time to consider participation.

Whānau will be provided a brief verbal summary of the study, including the main aims and processes of enrolment. Those that do consent to taking part in the study will be asked to sign consent forms. Those that do not consent will receive usual hospital care. At the time of consent, parent/s/caregivers will be asked if any other children in the household have been consented into the study. If so, the child will be allocated to the same intervention group (as outlined above). Participants will be informed they can withdraw their child, and/or themselves from the trial at any time, but that analysis of collected data will be on-going.

The intervention will consist of a 24-section tool with screening questions and follow-up protocols for each question in accordance with the process shown in Figure 2. Intervention group patients and their whānau are invited to go over the Harti tool with the RA and provided with an information and activity booklet (whānau booklet). After engagement, RAs will go through the Harti tool on a tablet. RAs will read out questions in order and fill in answers for participants, in some cases participants may view questions and select responses themselves, if they prefer. The RA will follow established protocols to follow up on any identified areas of need. For example, if a child is not enrolled with a particular service, the RA will offer to assist with enrolment; if the patient or another child present or at home is identified to have a sore throat, then Rheumatic Fever prevention protocols will be followed. RA's will have equipment (cell phones and an iPad) to enable referrals and enrolments to take place at the bedside or in the interview room. The Harti RA will be trained in delivering high quality smoking cessation interventions for whānau and safe sleep devices including Pēpi Pods will be made available. If the whānau are not sure of the child's enrolment status (anecdotally a common occurrence), the Harti RA will match the patient’s National Health Index (NHI) number with selected databases and double check for access or check enrolment with the service over the phone or directly, e.g., oral health services.

This group will receive usual hospital care and be offered the Marama satisfaction with care survey at the end of their hospital stay.

A summary of each research hypothesis is listed in Table 2 and the details of how these will be examined is below.

The primary endpoint is the relative risk of an acute hospital readmission in the 30 days following discharge for intervention group patients compared with control group patients (Hypothesis 1). The relative risk of readmission rather than readmission rate ratio was chosen as the primary endpoint because risk (probability of readmission within specified period) is easier to measure as it just requires counting the first time a patient is readmitted. By only counting the first readmission we avoid patients with large numbers of readmissions having a disproportion impact on the endpoint measure. This also simplifies statistical analysis. We have included the hazard ratio of readmission as a secondary endpoint (see below).

We will measure ethnic-specific differences in the primary endpoint measures and differences according to a 30-day² readmission risk for Māori and non-Māori patients (combined).

To determine the superiority limit for the purposes of sample size calculations it was considered that the Harti programme should deliver at least a 7% absolute decrease in overall readmission risk as this would be both clinically meaningful for tamariki and whānau and financially meaningful for the hospital. Furthermore, we have designed the study so that it has power to detect a 12% absolute decrease in the 30-day risk of readmission for the Māori population. It is not powered to detect differences in risk of readmission between Māori and non-Māori tamariki.

These will include the following: ➢

The hazard ratio for (first) readmission in the intervention group compared with the control group. This compares time to the first readmission rather than just the fact of a readmission. As such, it can make use of varying lengths of follow-up time and may be more sensitive to the impact of the intervention. However, the use of the hazard ratio relies on the proportional hazards assumption, which is that the relative hazard of readmission remains constant over time.

There are seven quantitative secondary endpoints. Pre-testing will be undertaken to ensure that data can be accessed (e.g., GP enrolment and WCTO records) for all study endpoints.

Hypothesis 2 examines the difference in satisfaction with hospital experience between the intervention and control groups. This will be measured using the first seven questions from the Mārama real-time survey. Two questions have been added for the purposes of this research: one on the responsiveness to whānau needs, the other on whether a holistic approach was employed (Figure 3). An tenth open-ended Mārama question will also be asked and analysed. The responses will be scored on a 5-point Likert Scale ranging from Strongly Disagree to Strongly Agree (1 point for the former, 5 for the latter), so the total possible score range for someone answering all questions is from 9 to 45 which will be scaled to a range of 0–100 to accommodate cases where one or more questions is not responded to. The responses will be provided by the child’s primary care-giver.

These hypotheses (3a & 3b) examine the difference in preventive services access and utilisation between the intervention and control groups. As previously mentioned, important universal preventative care is provided in the 0–4 year age group and for this reason these variables were chosen as secondary outcomes; general practice enrolment, oral health care, immunisations, and WCTO visits.

This endpoint has two parts: documentation and achievement. For documentation, external sources will be used to determine whether at admission the child was (i) enrolled with a General Practitioner (GP); (ii) enrolled with the oral health services; (iii) up to date with all oral health checks (seen in the last 12 months); (iv) up to date with all immunizations; (v) up to date with WCTO visits. The indicator that will be compared between the intervention and UC groups is the number of children with ‘non-compliance’³ in any of the five domains (GP enrolment, oral health enrolment, up-to-date oral health checks, up-to-date immunisation, and up to date WCTO visits). In whom compliance is achieved (fully documented in the intervention assessment or patient record during the child's admission), divided by the total number of children with non-compliance in any of these five domains.

For the achievement component of this endpoint (hypothesis 3b), which is contingent on detecting a significant difference in the compliance component, intervention and UC groups will be compared on the proportion of children who at admission were non-compliant on any of the five domains, and who became fully compliant within 30 days of discharge from the hospital. Fully compliant is defined as all of the following; enrolment with a GP, enrolment in oral health services, a completed or booked oral health check, complete on the pathway for immunisation (for age) and complete WCTO participation.

The immediate reduction of the number of people who smoke is of huge importance for Māori and one of the key health targets for the Ministry of Health NZ. Adult smoking rates for Māori are 21.3%, higher when compared with other ethnics groups such as Pacific (18.1%), and European/Other (7.9%) (35). It is clear that parental or household smoking increases children's risk of developing respiratory infections, which commonly result in hospital admissions (36).

For this endpoint (hypothesis 4), we will use the proportion of parent/caregivers where a referral to stop smoking services was made during the hospitalisation period for cessation services. This will be calculated by sending a list of parent/caregiver NHI numbers of all participant parent/caregivers to the sole face to face smoking cessation provider (Once and For All). The provider will then confirm if a referral had been received for that NHI number during the period of hospitalisation or within 30 days of discharge from hospital for the index admission.

It is well established that good housing is particularly important for the first 9-months of life (37). The Growing Up In New Zealand study showed that poor housing was strongly associated with adverse health outcomes, even when controlling for income (37). Healthy homes initiatives have been shown to reduce acute hospital admissions in a NZ setting (26). The Whare Ora programme is a Waikato Healthy Homes initiative that was developed in 2016 to help whānau to create healthier, warmer, safer and drier homes. The service is free to eligible whānau and provides healthy homes education, products such as curtains, heaters, dehumidifiers and draught stoppers, and referral of families for further services by other providers (38).

Hypothesis 5 examines the difference between intervention and UC groups in the proportion of children for whom a Whare Ora referral was made within 30 days post-discharge. This will be calculated simply as the number of Whare Ora referrals in each group divided by the total number of eligible households (as some participants live in the same household) in that group. Any household can be counted only once in the numerator.

Hypothesis 6 will compare differences between intervention and UC groups in median cumulative length of stay (LOS) in acute hospital admissions in the 12 months following discharge. Only coded events with lengths of stay of 3 h or more will be included in the calculation. Length of stay will be calculated in hours from the time of admission to the time of discharge using the hospital time stamp records. Length of stay will be calculated using, where appropriate, the rules established in the Ministry of Health's LOS indicator. These join events within the same districts where there is a transfer between the two and the second event commences less than 24 h after the end of the prior event. In these cases, if the initial admission is not acute then this and any length of stay in a joined subsequent hospitalisation, even if acute, will not be counted. Hospitalisations where none of the contributing stays is case-mix coded will also therefore be excluded, consistent with these rules. This endpoint will be calculated as the difference in median cumulative lengths of stay by individual (i.e., not summing for the group as a whole) and will include acute hospital stays at any hospital in the Waikato region. Using medians and individual level data will avoid outlier cases with very long lengths of stay having undue influence on the result of the test.

The frequency of readmission is explored in Hypothesis 7. This endpoint will be calculated as the rate ratio of acute readmission in the 12 months following discharge from the index admission event. It differs from the primary endpoint in using all admissions in the numerator, and a person-time denominator, which will exclude any days in which the patient is in hospital.

The only adverse event data collected will be alive status within 12 months of index discharge.

Analysis will be performed on an intention to treat basis. That is, endpoints will be compared for the two groups as defined from the point of randomisation to intervention or control group. For the primary endpoint, participants who are readmitted multiple times from each group will be included in the analysis only for the first (valid) occasion on which they are readmitted in the 12 months following the index admission.

Bivariate analyses comparing groups will test for statistical significance using the Chi Square test for differences in proportions. For the length of stay secondary endpoint where a difference in medians is to be tested, the Mann–Whitney U test will be used.

Although the randomised design is expected to yield balanced intervention and control groups in a study of this size, logistic regression analysis will be used for the primary and secondary endpoints to control for important confounding variables and to facilitate the assessment of interaction variables that increase or decrease the efficacy of the Harti tool. Statistical significance will be tested at the alpha level of 0.05. We will control for the false discovery rate in the secondary endpoints using the Benjamini–Hochberg method (39).

The study is not powered to detect ethnic differences in the impact of the intervention (effect modification). However, we will nevertheless conduct exploratory tests for effect modification with ethnicity (Māori/non-Māori), gender (male/female excluding other), and deprivation (as a dichotomised variable) using the Breslow Day test for homogeneity of odds ratio, and standard maximum likelihood methods (likelihood ratio tests) based on binomial regression models. We will assume an additive interaction for these analyses and report the relative excess risk due to interaction (RERI) where this is significant effect.

There is a large risk of contamination or a spill over effect—meaning the control group could get some of the intervention. This could happen for example if the nurses caring for control patients included some of the Harti tool screening questions in their assessments—more so than they normally would. The Hawthorne effect could also contaminate the project, meaning that just having this project happening may lead to better assessment and documentation for control patients. To mitigate the effect of contamination we will remove all Harti tool resources from the ward and stop it's use 6 months before the study starts. We hope this will give enough time for staff to get used to not doing a Harti assessment and to forget the individual questions; there is a high turnover of staff on the wards also so many working during the trial will have never used the Harti tool. We are working to ensure that Harti tool questions are not included into the design of new patient assessment forms. We will explain that a study is taking place but not go into detail for staff and we will not mention the Harti tool. Instead, we will emphasise the patient satisfaction component of the study. During the study period, the RA will keep all documentation, even the blank forms—confidential so as not to contaminate the ward environment.

To maintain blinding and concealment, the only person on the ward who will know explicitly what status the child has, is the RA. The child and whānau may be able to guess from the intervention but will not be told explicitly which group they are in. Results of the intervention assessment will be kept off the ward. As all patients will be consented and given the whānau satisfaction questionnaire by the RA, other ward staff will see the RA sitting with every admitted child and whānau. They may be able to tell who gets an intervention because the RA will spend longer with those who get the intervention than s/he does with those who are randomised to the control group and just get consented and the Mārama satisfaction with care survey. However, contamination is not expected to influence outcomes unduly. The following roles/steps will be blinded to treatment allocation; the person reviewing the notes of control and intervention patients to assess documentation of evidence of assessment, treatment/service delivery and referral/enrolment; the person matching NHIs with service data-bases to assess enrolment and service use; the person matching NHIs with hospital readmission data.

Information provided by participants will be only accessible to members of the research team. Participant study files and all other information will remain strictly confidential, unless there is an immediate risk of serious harm to them or others. Participant information will be logged into a study tracking sheet. No potentially identifying information will be used in any reports on this study. Records will be stored for at least 5 years in a secure place at Te Whatu Ora Waikato after the completion of the study. All electronic records will be password protected and stored on a restricted access shared drive at Te Whatu Ora Waikato.

The Health and Disability Ethics Committee has determined that a Data Safety Monitoring (DSM) Committee is not required.

The RCT protocol has ethical approval from the Health and Disability Ethics Committee (HDEC), reference number: 18/CEN/88 and is registered with the Australian New Zealand Clinical Trials Registry (ANZCTR), registration number: ACTRN12618001079235. The Harti study has locality approval from the Waikato District Health Board Māori Ethics Review Committee (Ref: RD017074).

In accordance with the He Pikinga Waiora Framework, we will engage and work with our networks and seek guidance from leaders, and academic and coal face experts in the Māori health, health, policy/Ministry of Health, iwi health, and political spheres to develop integrated knowledge mechanisms including dissemination approaches. Throughout the research process we will update key stakeholders through established networks; these include regional child health action groups, and paediatric and Māori networks. Traditional academic dissemination techniques will also be used including publication in peer-reviewed journals, and presentation at national and international paediatric, and Māori health conferences. We will keep local Māori stakeholders abreast of our process and have built the development of a comprehensive stakeholder engagement and dissemination plan into our timeline and budget.