The Conceptual Domain of Service Quality for Inpatient Nursing Services

Transcription

1 The Conceptual Domain of Service Quality for Inpatient Nursing Services Melissa M. Koerner HK&ASSOCIATES The purpose of this exploratory study is to broaden and deepen the conceptual domain of service quality for one service industry, inpatient nursing, by focusing on elements of service quality that have received insufficient attention in previous research. A conceptual definition of inpa- tient nursing service quality is provided, and an instrument based on that conceptualization is presented. The instrument s construct validity, nomological validity and reliability are ex- amined, and its relationship to SERVQUAL (Parasuraman, Zeithaml and Berry, 1985, 1988), an instrument that operationalizes the prevailing conceptual definition of service quality and has been used extensively in service quality research, is also investigated. The assumption underlying this study is not that the prevailing definition of service quality is incorrect, but rather, that is incomplete. Thus, the general intent of the Address correspondence to: Dr. M. M. Koerner, HK & Associates, 2299 Wyoming St., Salt Lake City, UT 84109, USA. This exploratory study examines the conceptual domain of service quality for inpatient nursing services. The findings suggest that the prevailing conceptual definition of service quality, as articulated by Parasuraman, Zeithaml, and Berry (1985, 1988), does not accurately describe service qual- ity for the customers of inpatient nursing services. A definition of inpatient nursing service quality is provided, and an instrument based on that defini- tion, the Inpatient Nursing Service Quality Scale, is presented. The instru- ment s construct validity, nomological validity, and reliability are examined, and the findings are favorable. The results of the study reveal that service quality perceptions for hospital inpatients consist of five dimensions: compas- sion, uncertainty reduction, reliability, close relationships, and individualized care. The dimensions are significant predictors of several outcomes, including global perceptions of service quality, willingness to recommend, and repurchase intentions. It is argued that industry-specific qualitative research should be conducted before using generic service quality measurement tools or instruments developed in other industries. J BUSN RES Elsevier Science Inc. All rights reserved. study is to identify service quality dimensions that supplement those that have already been described. What criteria do customers use to evaluate an organization s service quality? Parasuraman, Zeithaml, and Berry (1985, 1988) have developed a conceptual definition of service quality they believe answers this question for all types of services. Their definition is well-known, widely accepted, and viewed by both scholars and practitioners as an important contribution to the field of service quality. However, the definition lacks several elements that may be integral to the service quality construct in many service settings. These elements include the quality of interpersonal relationships between service pro- viders and those they serve (Berry, 1983; Brown and Swartz, 1989; Crosby, Evans, and Cowles, 1990; Gummesson, 1987), service provider effort (Mohr and Bitner, 1995a, 1995b), emotion and affect (Fineman, 1993; Gummesson, 1991; Hochs- child, 1983), social support (Adelman, Ahuvia, and Goodwin, 1994; Adelman and Ahuvia, 1995), and individualized service (Berry, 1995; Sasser and Fulmer, 1990). This study examines the conceptual domain of service quality for one type of service: inpatient nursing. Inpatient nursing was chosen, because it was believed that the emotional inten- sity of the inpatient experience and the strong interpersonal elements of the nurse patient relationship would highlight ser- vice quality themes that might otherwise go unnoticed (O Guinn and Faber, 1989; Schouten, 1991). In addition, satisfaction with inpatient nursing services has been shown to contribute more significantly than any other service to over-all perceptions of a hospital s service quality (Carman, 1990; Woodside, Frey, and Daly, 1989). The Prevailing Conceptualization of Service Quality Parasuraman, Zeithaml, and Berry (1985, 1988) (hereafter referred to as PZB) have undertaken a broad research agenda intended to define and measure the construct of service quality. In their initial research, they developed a series of propositions that were then integrated into a model, part of which Journal of Business Research 48, (2000) 2000 Elsevier Science Inc. All rights reserved. ISSN /00/$ see front matter 655 Avenue of the Americas, New York, NY PII S (98)

2 268 J Busn Res M. M. Koerner is called perceived quality, which represents the conceptual domain of service quality. To measure the perceived quality portion of the model, PZB developed the SERVQUAL scale, and they have assessed its reliability and validity repeatedly. Initially, 10 categories of service quality were identified. After subsequent testing, these categories were reduced to the following five dimensions. Reliability: the degree to which a promised service is per- formed dependably and accurately Assurance: the extent to which service providers are knowl- edgeable and courteous, and able to inspire trust and confidence Tangibles: the degree to which physical facilities, equipment, and appearance of personnel are adequate Responsiveness: the degree to which service providers are willing to help customers and provide prompt service Empathy: the extent to which customers are given caring, individualized attention In their research PZB consistently have found that customers rate reliability as the most important service quality dimension, regardless of the service studied. They also have found reliability to be the dimension with the greatest discrepancy between customers expectations and the service firm s perfor- mance. They conclude that The number one concern of customers today, regardless of type of service, is reliability (Zei- thaml, Parasuraman, and Berry, 1990, p. 28). Despite SERVQUAL s widespread acceptance and use, a number of criticisms have been brought against it. First, some researchers are skeptical that the dimensions apply to all ser- vice operations (Babakus and Boller, 1992; Cronin and Taylor, 1992; Dabholkar, Thorpe, and Renz, 1996). Otto and Ritchie (1995) found several common elements in consumers service experiences in five different service sectors, but they also observed significant differences in the nature and magnitude of the experiences across industries, with particularly striking differences between high- and low-involvement industries. Thus, it may not be possible for any instrument to measure service quality accurately in all industries. A second set of criticisms relate to the restricted view of service reflected in SERVQUAL. For example, Gummesson (1991) suggests that emotional components of service qual- ity compassion, sense of humor, and love have received fairly superficial treatment in service quality research, and that emotion should be measured in all service quality instruments, particularly those used in the helping professions. There is increasing evidence that emotional reactions to service en- counters are common and, in fact, represent the essence of the service experience (Otto and Ritchie, 1995, p. 54). In the PZB model, empathy is defined as caring, individualized service given to customers ; however, its operational definition, as reflected in SERVQUAL, emphasizes the personal attention rather than the caring aspect of the dimension. It seems that SERVQUAL addresses relatively superficial aspects of emotion. Similarly, individualized service identifying and responding to the customer s true needs in a way that acknowledges his or her individuality is viewed by several researchers as critical to service quality (Berry, 1995; Sasser and Fulmer, 1990, Surprenant and Solomon, 1987; Treacy and Wiersema, 1995), but it is not addressed fully in SERVQUAL. Individualized service is one aspect of PZB s empathy dimension, but rather than focusing on delivering service consistent with an individual s unique circumstances, the emphasis is on individual attention and personal attention. Again, it seems that SERVQUAL measures a relatively mild form of individualized service. Adelman, Ahuvia, and Goodwin (1994) suggest that social support, which consists of reducing uncertainty, improving self-esteem, and enhancing one s social connection to others, is an important aspect of service quality that has been overlooked. Social support has been found to increase customers willingness to recommend a service to others and to enhance over-all perceptions of service quality (Adelman and Ahuvia, 1995). Research in the nursing field suggests that one form of social support, regular and open information sharing between nurses and patients, is viewed positively by patients (Fosbinder, 1994; Ludwig-Beymer, Ryan, Johnson, Hennessy, Got- tuso, and Epsom, 1993; Morse, Bottorff, Anderson, O Brien, and Solbert, 1992). Social support is not addressed directly or indirectly in SERVQUAL. Service provider effort also may have an important impact on customers evaluations of service (Mohr and Bitner, 1995a, 1995b). Service provider competence and courtesy are addressed in SERVQUAL s assurance dimension, and service pro- vider promptness and helpfulness are assessed in the responsiveness dimension, but in rating these dimensions, the service provider s capability and effort are being evaluated simultaneously. Mohr and Bitner (1985) believe that effort may make a unique contribution to service quality evaluations; they found that effort had a positive influence on customers service evaluations, even when the service outcome was statistically controlled. In particular, service failures may be viewed quite differently, depending upon whether customers attribute the problem to factors within the service provider s control (i.e., effort), or outside their control (Folkes, 1987; Weiner, 1986). Finally, the quality of the relationship between service pro- viders and their customers is likely to be important in determining service quality perceptions (Dwyer, Schurr, and Oh, 1987), particularly in professional services (Berry, 1995; Grönroos, 1995). Bowen and Jones (1986) suggest that close, collaborative customer relationships are especially important when the service is characterized by high performance ambiguity and high buyer/seller goal congruity conditions that apply to many health-care service encounters. None of the items on the SERVQUAL scale addresses customer/provider relationships. In addition to concerns with SERVQUAL s conceptual framework, several researchers have been critical of SERV- QUAL s measurement properties, including its unstable factor structure, its use of difference scores in the analysis of results,

3 Service Quality for Inpatient Nursing Services J Busn Res 269 its use of self-reported importance weights, the validity, relia- to improve the service experience for both customers and bility and methodology of the scale, and ambiguity surrounding service providers. This goal is furthered when service quality the expectations section of the instrument (Baba- tools take both customer and service provider perspectives kus and Boller, 1992; Brown, Churchill, and Peter, 1993; into account and are used to pinpoint the similarities and Carman, 1990; Cronin and Taylor, 1992, 1994; Taylor, 1995; differences between customers and employees in their definitions Teas, 1993, 1994). However, despite these criticisms, SERV- of what constitutes high-quality service. Many research- QUAL and its derivatives (e.g., SERVPERF, Taylor, 1995) are ers imply that the service provider s opinion is irrelevant when still used extensively in service quality research and theory developing service measures, some even suggesting that the development (e.g., Crompton and MacKay, 1989; Cronin and only criteria that count in evaluating service quality are defined Taylor, 1994; Dabholkar, Thorpe, and Rentz, 1996; Dunlap, by customers (Zeithaml, Parasuraman, and Berry, 1990, Dotson, and Chambers, 1988; Finn and Lamb, 1991; Reiden- p. 16). However, when services are high in credence properties bach and Sandifer-Smallwood, 1990; Scardina, 1994; Spreng (as many medical services are), customers may not have the and Singh, 1993; Taylor, 1995; Woodside, Frey, and Daly, expertise to judge their quality at all (Darby and Karni, 1973). 1989). Interestingly, even the most ardent of these critics (e.g., Health-care providers quality judgements are not irrelevant Cronin and Taylor, 1992, p. 58) have indicated that the model when it comes to the health or survival of the patient. In fact, seems to define the domain of service quality adequately, research has shown that front-line employees have a unique and they have continued to conduct research based on that and valuable back stage perspective of service (Mangold conceptual domain. Although measurement issues are not and Babakus, 1991) and that the behaviors, perceptions, and the primary focus of this study, the growing concern about attitudes of service providers often are strikingly similar to SERVQUAL s measurement properties provides further impetus those of their customers (Parkington and Schneider, 1979; to explore alternative ways to define and measure service Schlesinger and Zornitsky, 1991; Schneider and Bowen, 1985; quality in different industries. Tornow and Wiley, 1991). Thus, accurate measures of service quality have the potential to generate data that improve service to customers, while also increasing the quality of work life The Need for Accurate for service providers. Measures of Service Quality Method Organizations need accurate measures of service quality to assure their continued survival and success. Research has shown repeatedly that service quality influences many impor- tant organizational outcomes. For example, service quality and customer satisfaction both have been found to be related to repurchase intentions (Bearden and Teel, 1983; Bolton and Drew, 1992; Cronin and Taylor, 1992; Oliver and Swan, 1989; Woodside, Frey, and Daly, 1989), and customers who rate service as excellent and particularly likely to intend to repurchase (Gale, 1994). Other behavioral intentions, such as word of mouth referrals and defections, also may be influ- enced by service quality (Zeithaml, Berry, and Parasuraman, 1996). Service encounter satisfaction has been linked to actual purchase behavior (LaBarbera and Mazursky, 1983), and service quality perceptions are related to willingness to recom- mend a company s service to others (Boulding, Kalra, Staelin, and Zeithaml, 1993; Parasuraman, Berry, and Zeithaml, 1991; Parasuraman, Zeithaml, and Berry, 1988). Gale (1994) has written extensively on the need for organizations to achieve market-perceived quality versus competitors. He argues that customer attraction and loyalty cannot be attained simply by providing superior service quality; customers must believe an organization s offerings are a superior value in relation to competitors offering. Any instrument that purports to measure service quality should be capable of predicting the organizational outcomes described above and should offer explicit guidance on which aspects of service must be addressed to achieve the outcomes. In addition, accurate measures of service quality are needed The study was conducted in two general phases. In phase one, qualitative research methods were used to explore the conceptual domain of service quality and clarify the service quality dimensions used by patients to evaluate an inpatient nursing experience. The qualitative findings of the study have been reported in detail previously (Koerner, 1996) and, therefore, are summarized only briefly here. In phase two of the study, scale development and testing, quantitative methods were used to develop an instrument to measure the dimensions identified in phase one and to assess the instrument s validity and reliability. The development and testing of this instrument, called the Inpatient Nursing Service Quality (INSQ) Scale, is the primary focus of this paper. Following a brief description of the qualitative research phase, the four stages in developing and testing the instrument are described. Both the qualitative and quantitative portions of the study were conducted through the hospitals of a major health-care system located in a large western community. Recent inpatients of the hospitals and nurses providing inpatient nursing services were the research subjects. Phase One: Qualitative Research to Conceptualize Inpatient Nursing Service Quality The qualitative research that preceded the development of the INSQ Scale consisted of three steps. First, an interdisciplinary literature review was conducted that focused on service quality in nursing and in other industries and service quality measure-

4 270 J Busn Res M. M. Koerner Table 1. Conceptual Definitions of Inpatient Nursing Service Quality Dimensions Compassion Positive experiences involved nurses who were perceived to demonstrate genuine compassion, defined as a sympathetic consciousness of each individual patient s vulnerability along with a desire to lessen it. Generally, more compassion was related to more positive experiences, and less compassion or indifference was related to more negative experiences. Preserving dignity Positive experiences involved nurses who treated patients respectfully and were concerned about helping them avoid embarrassment and preserve their sense of dignity. Generally, more concern for maintaining patient dignity was associated with more positive experiences. Close relationships Positive experiences involved nurses who were viewed as having ongoing, interpersonally close relationships with patients, including familylike relationships and friendships characterized by trust, liking, or love. In general, closer relationships occurred in more positive experiences, and more distant, professional, or businesslike relationships occured in more negative experiences. Individualized care Positive experiences involved nurses who were percieved to be sensitive to each individual patient s unique situation and needs, flexible and adaptable in their delivery of care, inclined to offer options to patients, able to anticipate patient needs, and able to make the best use of the patient s individual capabilities. Generally, more positive experiences were associated with more individualized service, and more negative experiences were associated with more standardized service. Uncertainty Reduction Positive experiences involved nurses who helped reduce patients uncertainty by teaching and explaining, reporting on the patient s progress and status, interpreting information from doctors and providing additional support. Generally, more uncertainty reduction was related to more positive experiences and less uncertainty reduction was related to more negative experiences. Extra effort Positive experiences involved nurses who were perceived as psychologically engaged in their interactions with patients and trying hard, which included working with extra intensity, putting in extra time, taking unusual risks, performing extra-role activities, and displaying extra responsiveness. In general, high effort was related to more positive experiences and average or low effort was related to negative experiences. ment. On the basis of those findings, an interview protocol dures, the nurses and patients were asked to describe in detail was developed and used to conduct depth interviews with one or two of their most positive patient care experiences and nurses and patients. one or two of their most negative patient care experiences. To recruit informants for the qualitative phase of the study, The interviews were audiotaped and transcribed and then an announcement memo was sent from the hospital adminis- were analyzed using Strauss and Corbin s (1990) constant trator to the heads of all nursing departments. Managers were comparative method. According to this method s procedures, urged to discuss the study with their nurses and request interview transcripts were first examined closely and open volunteers. Two or three nurses were recruited from each coding was used to name and categorize phenomena. At the nursing unit, including intensive care/critical care, intermedi- conclusion of this step, the major themes that emerged in ate care, pediatrics, obstetrics gynecology, orthopedics/urology, the interviews were identified (e.g., receiving compassionate and medical/surgical. Those who volunteered for the care and being embarrassed ), and examples of each theme, study participated in a brief telephone interview with the taken from the interviews, were compiled. Next, the tran- researcher to determine their suitability for the study. In this scripts were re-read and axial coding was used to examine the interview, each nurse was asked to identify a patient with context and conditions under which the phenomena occurred whom he or she had a recent positive experience, and this and the causes and consequences of the phenomena. This patient was also contacted for an interview. In all, 15 nurses step served to clarify how the themes related to each other and 14 of their patients were interviewed. The nurses ranged and to very positive and very negative service experiences. in age from 25 to 55, and all were Caucasian. Two were males For example, positive experiences were associated with indi- and 13 were females, and their nursing experience ranged vidualized care and close personal relationships with nurses; from 2 to 35 years. The patients ranged in age from 16 to whereas, negative experiences were associated with standard- 74, and all were Caucasian. Nine were female and 5 were ized care and businesslike or distant relationships with nurses. male. The patients had been hospitalized for a wide variety Finally, selective coding was used to assemble the findings into of reasons, including childbirth, knee surgery, cancer, heart an emergent model, which included conceptual definitions attack, meningitis, head trauma, and back injury. for the six service quality dimensions that were identified in The interviews were conducted within one hospital owned the analysis. The themes that emerged in nurses and patients and operated by the health-care system sponsoring the research. interviews were very similar, so they were integrated into one Using McCracken s (1988) depth interviewing proce- model. The service quality dimensions identified in phase

5 Service Quality for Inpatient Nursing Services J Busn Res 271 Table 2. Phase Two: INSQ Scale Development and Testing Procedures Stage 1: Domain specification Refine conceptual definitions of dimensions Generate scale items Review and pretest items Stage 2: Data collection Design survey administration procedures Select sample Administer survey Stage 3: Development of measurement model Step 1: Perform exploratory factor analysis (SERVQUAL, New Items) Determine dimensions (INSQ) Modify instrument Step 2: Perform confirmatory factor analysis Assess convergent and discriminant validity and reliability Refine instrument Stage 4: Assessment of nomological validity Examine relationships among dimensions and service quality constructs cluded in the instrument (e.g., The nurses were consistently courteous with me and The nurses gave me prompt care ). The decision to use only the perceptions section of SERVQUAL (rather than both the expectations and perceptions sections) was made because of the growing concern among researchers about SERVQUAL s use of difference scores and explicit expectations ratings, and in light of compelling evidence that the perceptions section alone is a superior predictor of service quality (Taylor, 1995). Five items were included on the instrument to measure dependent variables commonly used in service quality research: global perception of nursing service quality (two items), perception of service quality relative to competitors (one item), intentions to repurchase the service (one item), and willingness to recommend the service to others (one item). The instrument was pretested with five former patients (nonacademic associates of the researcher who had experienced a hospitalization) to measure completion times, determine areas of confusion, and assess affective responses to the survey. Several minor modifications were made based on the pretesting. one of the research are compassion, preserving dignity, close STAGE 2: DATA COLLECTION. The sample frame for data col- relationships, individualized care, uncertainty reduction, and extra lection consisted of patients who had been discharged from effort. The conceptual definitions for these dimensions are the sponsoring organization s inpatient medical facilities dur- shown in Table 1. ing the month before the survey was administered. To select participants from the sample frame the following procedure Phase Two: Scale Development and Testing was used. The sponsoring organization conducted a brief tele- As Briggs and Cheek (1986, p. 142) and others have argued, phone survey with all former inpatients within 10 days of thorough conceptual analysis should precede data collection. their discharge from its hospitals. Over a 2-week period, at The purpose of phase one, therefore, was to identify and the end of the telephone survey, respondents were asked if conceptualize the new service quality dimensions; phase two they would be willing to complete a written survey focusing was intended to create and test a tool to measure those dimen- on the nursing care they received during their hospital stay. sions. INSQ Scale development and testing occurred in four Four-hundred-eighteen volunteers were identified through stages, which are summarized in Table 2. this process and were sent the survey the following week. Approximately 4 weeks after volunteers were solicited for the STAGE 1: DOMAIN SPECIFICATION. The objective of this stage study, 249 questionnaires had been returned. This represents was to specify the definition for each new service quality a response rate of 60% of those who received the questiondimension further. The researcher accomplished this by writ- naire, and approximately 21% of those who were asked during ing survey items to operationalize each dimension, being care- the telephone survey to participate in the mail survey. ful to follow closely the conceptual definitions. During this The demographic characteristics of respondents are as folprocess, care was taken to use simplified language, and when- lows. Women were over-represented in the study: 69% of the ever possible, words and phrases used by the patients them- respondents were female, and 31% were male. Participants selves during the interviews were included in the questions were fairly evenly distributed across age categories: 17% were (e.g., The nurses genuinely cared about me and The nurses between the ages of 18 and 25; 36% were between 26 and really understood my personal situation ). 45; 20% were between 46 and 64, and 25% were over 64. A pool of between 5 and 15 survey items was developed Respondents had been hospitalized at six different hospitals for each dimension, with the goal of having 4 to 8 items per in one state, ranging from small, community-based hospitals dimension in the final version of the instrument (Bagozzi, to large tertiary care hospitals. 1994). The items were reviewed by two individuals with survey development experience, four experienced professors of STAGE 3: DEVELOPMENT OF MEASUREMENT MODEL marketing and management, and the sponsoring organizathe Exploratory Factor Analysis. The first step in developing tion s research department staff. Survey items were modified measurement model was to examine several potentially based on their feedback. meaningful structures in the data collected from former pa- In addition to the items generated for this study, the 22 tients. Exploratory factor analysis (EFA) is commonly viewed items from the perceptions section of SERVQUAL were in- as the best analytical tool for this purpose (Briggs and Cheek,

6 272 J Busn Res M. M. Koerner Table 3. Factor Matrix for SERVQUAL Items, After EFA Factor Item Factor The nurses had my best interests at heart. (empathy) The nurses gave me personal attention. (empathy) When I had a problem, the nurses showed a sincere interest in solving it. (reliability) The nurses understood my spacific needs. (empathy) The nurses were always willing to help me. (responsiveness) The nurses gave me individual attention. (empathy) The behavior of the nurses instilled confidence. (assurance) The nurses were never too busy to respond to my requests. (responsiveness) The nurses appeared neat. (tangibles) The nurses were consistantly courteous with me. (assurance) The nurses had the knowledge to answer my questions. (assurance) Factor When the nurses said they would do something by a certain time, they did it. (reliability) The nurses provided care at the time they said they would. (reliability) The nurses gave me prompt care. (responsiveness) Factor The hospitial had modern-looking equipment. (tangibles) The hospitial s facilities were visually appealing. (tangibles) I felt safe in the environment at the hospital. (empathy) The nurses performed treatment and care right the first time. (reliability) Factor The written information I received from the nurses about my medical condition was visually appealing. (tangibles) The nurses told me exactly when procedures would be performed. (reliability) Factor The hospital had visiting hours convenient to all its patients and their families. (empathy) Eigenvalue Percentage of variance explained Cumulative percentage of variance explained 1986); therefore, a series of EFAs were performed. For each one factor contained items from each of the five SERVQUAL EFA, principal components analysis and orthogonal rotation dimensions, another factor contained a combination of reliabilwere used because the objectives of the analysis were: (1) to ity and responsiveness items, and a third factor contained items summarize the data into a minimum number of factors; (2) from the tangibles dimension. The two remaining factors con- to identify relatively discrete factors; and (3) to use the factors tained two items and one item, respectively, but were not for prediction purposes (Hair, 1995). In addition, for each interpretable. As shown in Table 3, the eigenvalue for the EFA, several criteria were used in determining the number of first factor was 12.44, representing 59% of the variance. The factors to extract (e.g., eigenvalues and scree tests), rather second, third, fourth, and fifth factors eigenvalues were 1.41, than using the eigenvalue-one procedure, which can produce 0.98, 0.91, and 0.68, respectively. The five factors represented distorted results if used in an arbitrary manner (Comrey, 1978). a total of 78% of the variance in the SERVQUAL variables. Because the items developed for this study were intended Second, EFA was performed on the service quality items to supplement, rather than replace, the SERVQUAL scale, EFA developed for this study (excluding the SERVQUAL items) in was first performed on the SERVQUAL items to determine an iterative fashion to determine their dimensionality. Based that scale s dimensionality. In light of the extensive previous on the qualitative research phase of the study, a first stage testing of SERVQUAL s five-factor structure, it was decided conception of the factor structure (Comrey, 1978) was formulated: to extract five factors. Table 3 shows the SERVQUAL items six factors were expected to be produced. However, and their factor loadings, after EFA. The SERVQUAL items after examining a variety of solutions, the most meaningful did not load with their assigned dimensions. Of the five factors, produced four distinct factors, three of which were consistent

7 Service Quality for Inpatient Nursing Services J Busn Res 273 Table 4. Factor Matrix for INSQ Scale, After EFA Factor Item Compassion The nurses treated me with respect I had the nurses full attention when they were with me The nurses genuinely cared about me The nurses helped me keep my sense of dignity during my hospital stay The nurses were kind to me. Uncertainty Reduction The nurses gave me and my family members frequent updates on my condition The nurses helped me understand information given to me by my doctor The nurses regularly explained what was or would be happening to me during my hospital stay The nurses regularly checked with me to see if I had any special concerns or questions The nurses sometimes knew what I wanted or needed even before I asked for it. Reliability When the nurses said they would do something by a certain time, they did it The nurses gave me prompt care The nurses provided care at the time they said they would The nurses were never too busy to respond to my requests. Close Relationships The nurses and I talked about things in our lives other than my medical concerns The nurses and I sometimes kidded, laughed, or joked with each other The nurses and I enjoyed each other s company The nurses and I liked each other. Individualized Care The nurses took my unique situation into account in caring for me The nurses were willing to do things a little differently for me if my situation required it The nurses really understood my personal situation The nurses knew my individual preferences and needs The nurses made sure I understood any instructions I was given Eigenvalue Percentage of variance explained Cumulative percentage of variance explained with hypothesized dimensions: close relationships, uncertainty reduction, and individualized care. Items from the dimensions of compassion and dignity consistently loaded together and were combined to form a single dimension. Items from the effort dimension produced dominant loadings on all four dimensions, so effort was disregarded as a distinct dimension. The eigenvalues for the four factors were 10.83, 1.26, 1.05, and 0.75, and explained 57, 7, 6, and 4% of the variance, respectively. In total, the four factors represented 73% of the variance in the new items developed for this scale. Third, the service quality items developed for this study were combined with SERVQUAL to form the Inpatient Nursing Service Quality (INSQ) Scale, and EFA was performed on these items. The original intent was to include all of the SERVQUAL dimensions in this procedure. However, after inspecting the SERVQUAL factor matrix, it was decided to incorporate only items from SERVQUAL s second factor into the INSQ Scale, because this was the only dimension that was clearly interpretable and not already represented by the other INSQ dimensions. All of the items in this factor were reliability/ responsiveness items. The factor matrix for the INSQ Scale, after the EFA, is shown in Table 4. The eigenvalues for the INSQ Scale factors were 12.71, 1.53, 1.27, 1.00, and 0.74, and represent 55, 7, 6, 4, and 3% of the variance, respectively. In all, 75% of the variance in INSQ items is explained by the five factors. After completing the EFA for the INSQ Scale, the remaining service quality items produced primary loadings on a single factor and secondary loadings differing by 0.10 or more. Five distinct factors emerged: compassion, close relationships, uncertainty reduction, individualized care, and reliability. Confirmatory Factor Analysis. Once a clear 5-factor structure had been identified for the INSQ, the solution was subjected to CFA using Lisrel 7 (Jøreskog and Sorbom, 1989). Lisrel

8 274 J Busn Res M. M. Koerner Table 5. Goodness-of-Fit Indices for SERVQUAL and INSQ Scale Model n x 2 df GFI AGFI RMR INSQ Scale Null model * Factor model * SERVQUAL Null model * Factor model * GFI goodness-of-fit index; AGFI adjusted goodness-of-fit index; RMR root mean squared residual. * p was used because of its ability to provide detailed diagnostic Convergent Validity. Convergent validity of the INSQ Scale information about a measure s reliability and validity, includ- was assessed in five ways. First, the goodness-of-fit indices ing the degree of model fit, data regarding convergent and are above or approaching 0.90, which in intself is an indication discriminant validity, and information about method and error of convergent validity (Bagozzi and Yi, 1988). Second, all of variance (Bagozzi, Yi, and Phillips, 1991). the individual items have a statistically significant factor load- During the CFA, several items were deleted from the INSQ ing on their assigned dimensions as indicated by the factor Scale as indicated by the modification indices. In the analysis, loadings and t-values listed in Table 6 (Anderson and Gerbing, items were permitted to load only on their assigned factors, 1988). Third, as was seen in Table 4, the secondary loadings with cross-loadings set to zero, and the intercorrelations among on factors to which items are not assigned are not large. Fourth, the factors were freely estimated. The covariance matrix for Table 6 indicates that the average proportion of variance exthe items was used in the analysis, and parameter estimates plained by each dimension is 0.59 or higher, which exceeds were made under the maximum-likelihood method. For com- Bagozzi and Yi s (1988) criterion of 0.5 or higher. Therefore, parison purposes, the analysis also included a null model (no the dimensions specified in the INSQ Scale measurement constructs were recognized among the observed variables). model seem to account for a substantial proportion of variance The results of this analysis are shown in Table 5. in the items used to measure them. These findings suggest The fit tests that are relatively less dependent on sample that convergent validity is established for the INSQ Scale. size suggest a good fit for the 5-factor model (GFI 0.917, Discriminant Validity. Discriminant validity was assessed in AGFI 0.870), although the chi-square tests for both the three ways. First, the correlation between each pair of dimennull and 5-factor models of the combined scale are significant, sions was examined to determine if it is significantly different suggesting an unsatisfactory fit ( , 91 df for the than 1.0 (Schmitt and Stults, 1986). Although this test is not null model, and 2 114, 67 df for the 5-factor model). a rigorous one, Table 7 shows that the dimensions are not Nevertheless, the 5-factor model produced a chi-square statishighly correlated. A similar procedure for assessing discrimitic of less than twice the degrees of freedom, which is a comnant validity is to determine if the covariance plus two stanmonly accepted method of assessing fit (Podsakoff and MacKendard errors for each pair of dimensions add to less than 1.0 zie, 1994). In addition, the difference between the chi-square statistic in the 5-factor model does show significant improveduced values ranging from 0.36 to 0.63 for each pair of (Dabholkar, Thorpe, and Rentz, 1996). This procedure pro- ment over the null model ( d2 1729, 24 df; GFI 0.70). The complete SERVQUAL scale also was subjected to CFA, dimensions, suggesting that the INSQ dimensions are distinct with items assigned to their hypothesized dimensions. Again, even when measurement error is taken into consideration. items in the 5-factor SERVQUAL scale were permitted to load A third, more stringent method of assessing discriminant only on their assigned factors, with cross loadings set to zero, validity has been outlined by Fornell and Larcker (1981). and the intercorrelations among the factors were freely esticonstruct among the individual items included in the con- They contend that the average variance accounted for by the mated. The covariance matrix for the items, again, was used in the analysis, and parameter estimates were made under the struct should be greater than the amount of variance the maximum-likelihood method. For the SERVQUAL scale, the construct shares with any other construct. Satisfying this crite- analysis indicated a poor fit for the null model ( , rion shows that the measures within the dimension have more 210 df;gfi 0.131, AGFI 0.044) as well as for the 5-factor in common with each other than the dimension has with other model ( 2 767, 179 df; GFI 0.79, AGFI 0.63). For dimensions. Table 6 shows the average variance extracted for the 5-factor model the 2 is more than twice its degrees of each dimension, and Table 7 shows the variance shared by freedom, again, suggesting a poor fit, although there is imthe each pair of dimensions. Again, in every case, the square of provement in the 5-factor model over the null model (x 2 construct intercorrelations is less than the average variance 2872, 31 df; GFI 0.58). extracted for items in each dimension. These findings provide

9 Service Quality for Inpatient Nursing Services J Busn Res 275 Table 6. Pattern Coefficients, Reliabilities and t-values for INSQ Scale Items Individual Average Pattern Item Variance Coefficient* Reliability Extracted t-value Item Description 0.60 Compassion 0.81 (0.050) I had the nurses full attention when they were with me (0.050) The nurses genuinely cared about me (0.058) The nurses were kind to me Uncertainty Reduction 0.75 (0.055) The nurses helped me understand information given to me by my doctor (0.057) The nurses regularly explained what was or would be happening to me during my hospital stay (0.048) The nurses regularly checked with me to see if I had any special concerns or questions (0.052) The nurse sometimes knew what I wanted or needed even before I asked for it Reliability 0.94 (0.047) The nurses gave me prompt care (0.047) The nurses provided care at the time they said they would Individualized Care 0.80 (0.047) The nurses took my unique situation into account in caring for me (0.041) The nurses really understood my personal situation (0.050) The nurses knew my individual preferences and needs Close Relationships 0.70 (0.065) The nurses and I sometimes kidded, laughed or joked with each other (0.062) The nurses and I enjoyed each other s company. Standard errors are listed in parentheses following the factor loadings. * Pattern coefficients represent the relationship between the observed indicators (items) and the latent constructs (dimensions). evidence for the discriminant validity of dimension in the pair of dimensions at a time. For each pair of dimensions, INSQ Scale. this procedure resulted in significantly lower chi-square values As a final assessment of discriminant validity, the phi matrix for the unconstrained models, with differences that exceeded for each pair of dimensions was fixed at 1.0, and then freed, the critical chi-square value at p 0.01 in every case. Again, and chi-square difference tests were performed to determine this provides substantial evidence for the discriminant validity whether the values for the unconstrained models were significantly of the dimensions. lower than those of the constrained models (Anderson Reliability. Individual item reliabilities are listed in Table 6. and Gerbing, 1988). This procedure was performed for one All but one are above 0.5, which exceeds Bagozzi and Yi s Table 7. Means, Standard Deviations, Alphas and Intercorrelations for INSQ Scale Dimensions r Dimension M SD Compassion Individualized care (0.48) Reliability (0.40) 0.66 (0.44) Relationships (0.36) 0.69 (0.48) 0.48 (0.23) Uncertainty reduction (0.49) 0.75 (0.56) 0.62 (0.38) 0.60 (0.36) 1.0 Nursing service quality M mean; SD standard deviation; coefficient alpha. The square of the construct correlation (variance shared) is listed in parentheses after correlation coefficients.

10 276 J Busn Res M. M. Koerner Table 8. Summary of Initial Nomological Validity Assessment Nursing Service Willingness To Intent To Service Quality Dimension Quality Recommend Repurchase Relative to Competitors Compassion 0.75* 0.71* 0.63* 0.62* Individualized care 0.72* 0.62* 0.55* 0.61* Reliability 0.70* 0.62* 0.55* 0.51* Close relationships 0.63* 0.52* 0.43* 0.47* Uncertainty reduction 0.77* 0.63* 0.55* 0.57* Note: For all dimensions, Pearson correlations are reported. * Significant (1988) criterion. Additionally, INSQ scale reliability was as- variance in over-all quality of nursing is explained by the sessed by calculating the internal consistency reliability (Cron- INSQ dimensions. The t-values for all of the dimensions except bach, 1951) of the items included in each of the dimensions. individualized care are significant (p 0.01). Uncertainty reduc- As shown in Table 7, Cronbach s alphas ranged from 0.74 to tion, reliability, and compassion are the best predictors in the 0.91, suggesting highly reliable scales. equation, followed by close relationships. This finding contradicts PZB s assertion that reliability is the most significant STAGE 4: ASSESSMENT OF NOMOLOGICAL VALIDITY. To examcontributor to over-all service quality perceptions. The finding ine the nomological validity of the INSQ Scale, several analyses also shows that the INSQ dimensions do contribute to global were conducted to investigate the nature of the instrument service quality perceptions, which is another indication of the and its relationship to several other service quality constructs. scale s nomological validity. Of primary interest were four dependent variables commonly Second, service quality relative to competitors was regressed used in service quality research: over-all perceptions of nursing on the INSQ dimensions. Although the equation is significant quality, perceptions of hospital service quality relative to com- (43% of the variance in service quality is accounted for by petitors, willingness to recommend the hospital to others, and the dimensions), only compassion and individualized care are intention to repurchase the hospital s service. significant predictors. Third, willingness to recommend was regressed on the INSQ Relationships Among INSQ Scale dimensions. Again, the equation is significant; 55% of the Dimensions and Dependent Variables variance in willingness to recommend is accounted for by the As an initial assessment of the relationships among the INSQ INSQ dimensions. In this equation, however, compassion and dimensions and each of the dependent variables, correlation reliability are the only significant predictors. analysis was performed. Table 8 shows that each of the dimen- Finally, intent to repurchase was regressed on the INSQ sions is positively and significantly related to the dependent dimensions. Again, the equation is significant, 43% of the variables. This outcome was expected in light of previous variance in intent to repurchase is explained by the INSQ research suggesting that favorable perceptions of important dimensions. In this equation only compassion and reliability aspects of service quality lead to willingness to recommend are significant predictors. and intent to repurchase the service. The finding provides In summary, the results of the regression analyses provide evidence of nomological validity of the INSQ Scale. further support for the nomological validity of the INSQ Scale. Another preliminary assessment involved examining the correlation between the INSQ Scale s 14 items and SERV- QUAL s 21 items, with all items equally weighted and aver- Discussion aged. (It should be noted that the INSQ Scale contains two items that are also used in SERVQUAL). The correlation be- Inpatient Nursing Service Quality Dimensions tween the two scales is high (0.90) and demonstrates substanbut The major finding of the study is that five dimensions all tial nomological validity of the INSQ. one different from the dimensions included PZB s service Next, a series of regression analyses were conducted to quality model account for a substantial proportion of varidetermine the extent to which INSQ dimensions predict val- ance in patient perceptions of inpatient nursing service quality. ued organizational outcomes. For each of these analyses, all Of these five dimensions, uncertainty reduction has the greatest five INSQ dimensions were entered into the equation, using influence on nursing service quality perceptions, followed SPSS PC. The detailed results of these analyses are presented closely by reliability and compassion, and then, by close relationin Table 9, and summarized in Table 10. ships. Individualized care is a significant predictor of service First, the over-all nursing quality scale was regressed on the quality relative to competitors despite the fact that it does INSQ dimensions. The equation is significant; 70% of the not make a unique contribution to nursing service quality