Development and validation of Malaysian one stop crisis center service quality instrument (OSCC-Qual) for domestic violence management

A major challenge faced by survivors of domestic violence (DV) is the onerous task of seeking services from various agencies [1]. As these agencies are often located in different places, this entails the necessity for the survivors to move from place to place. For example, the survivors may need to go to the hospital to seek treatment for their physical injuries, to the police station to make police reports and in some instances, to meet the social workers to obtain temporary shelter. To overcome these problems, an integrated center known as One Stop Crisis Center (OSCC) was first set up in Malaysia back in 1994 and subsequently established in many parts of Southeast Asia and the Western Pacific regions [1]. OSCC can be defined as “an inter-professional, health-system based center that provides survivor-centered health services alongside some combination of social, legal, police and/or shelter services to the survivors” [1].

Although OSCC has been established for almost three decades, there is a lack of literature on service quality measurement in OSCC. Unfortunately, healthcare services can be highly complex [2] and bureaucratic [3]. What is perceived as a quality service by one stakeholder might be perceived differently by another; thus, making it difficult to have a “one-size-fits-all” service quality instrument in the healthcare sector using generic service quality instruments such SERVQUAL, HEALTHQUAL, SERVPERF, PubHosQual and HospitalQual [2, 4].

Colombini et al. [1, 5] constructed a five-pronged healthcare recommendations framework in the management of cases in OSCC. These five recommendations are that (1) healthcare providers need to have good knowledge and awareness about domestic violence, protocols and referral networks to manage these cases; (2) healthcare providers need to have the skills and competency to examine and manage injuries sustained by the survivors; (3) healthcare providers need to have the right attitudes (e.g., non-judgmental and non-condescending attitude in accepting survivors as who they are and demonstrating empathy); (4) the need to have a conducive healthcare environment (i.e., sufficient time allowed for proper enquiry and response) and (5) healthcare providers need integrity and good ethical principles (e.g., keeping confidentiality, maintaining privacy, being respectful and prioritizing the survivor’s safety). Although a number of guidelines and instruments on training and integrating DV responses in health centers have been published [6‐7]; to the best of our knowledge, there is a lack of instrument on measuring the quality of healthcare services rendered in the Malaysian OSCC from the perspective of DV survivors. By using Colombini et al. [5] as our overarching conceptual framework, we developed and validated a new service quality instrument to measure service quality in Malaysian OSCCs using a sequential process of item development, instrument development and instrument validation [8].

This study was divided into 5 stages. In stage 1, group discussions were conducted to identify potential items measuring service quality in OSCC based on Colombini et al. [5]. This was achieved through group consensus. In the event that there are discrepancies or disagreement, further discussions would be held with other emergency physicians with experience handling DV cases in OSCC in Malaysia and who were not part of this study. Apart from utilizing Colombini et al. [5], two healthcare service quality instruments that most closely reflect services in OSCC, i.e., Hong Kong Inpatient Experience Questionnaire (HKIEQ) by Wong et al. [9] as well as Rakhmawati et al. [10], were also referenced to identify the potentially relevant items. HKIEQ is a service quality instrument that measures in-patient experience in 9 dimensions, i.e., prompt access, information provision, care and involvement in decision-making, physical and emotional needs, coordination of care, respect and privacy, environment and facilities, handling patient feedback and overall care of health professionals and quality of care [9]; whereas the service quality instrument for public health center by Rakhmawati et al. [10] measures 4 dimensions, i.e., quality of healthcare service delivery, quality of healthcare personnel, quality of administration process and the adequacy of healthcare resources.

In stage 2, content validation was conducted using content validity index (CVI) and modified kappa [11, 12] to determine the relevance and appropriateness of an item. CVI is defined as the proportion of content experts who rate the relevance of an item with scores of 3 or 4 on a 4-point Likert scale (where 1 = not relevant at all, 2 = somewhat relevant, 3 = quite relevant, and 4 = highly relevant) [11]. A CVI value of 0.85 and above was considered as valid (Lynn 1986). However, to account for the possibility of chance agreement in CVI, modified kappa (κ) was analyzed (with the criteria that if κ = 0.40–0.59, the inter-rater agreement is interpreted as “fair”; if κ = 0.60–0.74, it is interpreted as “good” and if κ > 0.74, it is interpreted as “excellent”) [12]. Based on the recommendations by Lynn [11] a minimum of ten experts were required. In this study, 13 experts who had experience in managing OSCC cases in Sarawak General Hospital participated in this stage. These experts consisted of one senior consultant and head of the emergency and trauma department of Sarawak General Hospital, 9 emergency physicians, 1 nursing matron and 2 nursing sisters evaluated the relevance of items measuring service quality in OSCC.

In stage 3, exploratory factor analysis using IBM SPSS Statistics software version 25 for Mac was conducted to identify the number of factors or constructs to be extracted as well as items with good validity. Principal component analysis with varimax rotation was used as the extraction method. Scree plotting was performed with eigenvalue of > 1.0 used as the cut-off value to determine the numbers of factors. Factor loading > 0.4 was used as the criteria to determine whether an item is to be included or removed [13]. With regards to the sample size needed for this stage, the guideline for subject-to-item ratios by Costello & Osborne [14] was used. Most of the studies described in Costello & Osborne [14] used subject-to-item ratios that ranged from 3:1 to 10:1. In this case, due to a shortage of healthcare staff in the emergency department of Sarawak General Hospital, a 3:1 subject-to-item ratio was adopted, by recruiting 141 healthcare staff (medical doctors, assistant medical officers, and staff nurses) who had at least 2 years’ experience handling OSCC cases. These participants were recruited through convenience sampling. Participants then evaluated the items on a written questionnaire using a five-point Likert scale, where “1 = strongly disagree” to “5 = strongly agree”. House officers or interns as well as healthcare staff with less than 2 years’ experience in handling with OSCC cases were excluded. The reason why healthcare providers were selected as participants for stages 2 and 3 rather than patients (in this case, DV survivors) is because patients often lack the necessary knowledge to reliably list and evaluate the technical aspects of healthcare services quality, such as judging a doctor’s skills.

In stage 4, confirmatory factor analysis was conducted to further ascertain the validity of the items and factors or constructs retained from the previous stage. The measurement modelling of Partial Least Square Structural Equation Modelling using SmartPLS version 3.0 [15] was used in this stage. Cronbach alpha, composite reliability index and the rho A (ρA) coefficient (Dijkstra-Henseler’s rho) [16] were performed to determine internal consistency. For convergent validity, item factor loadings and the Average Variance Extracted (AVE) values of the constructs were determined. AVE refers to the grand mean value of the squared loadings of all items associated with a factor. Factor loading of > 0.7 is considered as acceptable for inclusion, whereas factor loading of < 0.4 would be removed [17]. For item with factor loading between 0.4 and 0.7, AVE would then be considered. If the AVE > 0.5, the item would be included [17]. For discriminant validity, the Fornell and Larcker criterion [18] and potential cross loading were evaluated. Fornell and Larcker criterion measures the degree to which an item loads higher on its own construct or factor (as measured using the square root of its AVE value) in relation to its correlation with other constructs or factors (as measured using the square of correlation values) [18]. In this stage, DV survivors who were managed in OSCC in Sarawak General Hospital were recruited as our participants. Using the F-test in G*Power software version 3.1.9.3 for Mac, with the criteria to achieve a minimum effect size of f² = 0.15; α error probability = 0.05, power level (1-β) = 0.8 and with 7 independent factors, 110 participants were recruited (comprising of 92 female and 18 male participants with age ranged from 18 to 70 years old, mean = 36.8 years old). Survivors with severe injuries or those with hemodynamic instability requiring urgent medical intervention were excluded. The participants rated the items using written questionnaire with a five-point Likert scale, where “1 = strongly disagree” to “5 = strongly agree”.

In stage 5, translation into local languages that are familiar to the Malaysian communities were conducted. The translation processes outlined by Beaton et al. [19], i.e., (1) forward translation (2) synthesis and harmonization of forward translated versions (3) back translation and (4) review and resolution of discrepancies (stage 5). Professional translation service was utilized for the forward and backward translation.

The medical research ethics approval from the Malaysian Medical Research and Ethics Committee (NMRR-20-1437-5483; https://​nmrr.​gov.​my/​) was obtained before starting this study. No personal identifiable information such as the participants’ names, national identity number or passport number, etc. were collected in this study. All participants in stages 2 to 4 were assured that their data were used solely for the purpose of this study and not for other purposes.