Oliveira Lopes Marcos - Clinical Research Designs in Cardiology and the Use of Nursing Taxonomies

Transcription

1 QCVC Committees Scientific Activities Central Hall General Information FAC Thematic Units Arrhythmias and Electrophysiology Basic Research Bioengineering and Medical Informatics Cardiac Surgical Intensive Care Cardiomyopathies Cardiovascular Nursing Cardiovascular Pharmacology Cardiovascular Surgery Chagas Disease Echocardiography Epidemiology and Cardiovascular Prevention Heart Failure Hemodynamics - Cardiovascular Interventions High Blood Pressure Ischemic Heart Disease Nuclear Cardiology Pediatric Cardiology Peripheral and Cerebral Vascular Diseases Sports Cardiology Technicians in Cardiology Transdisciplinary Cardiology and Mental Health in Cardiology Clinical Research Designs in Cardiology and the Use of Nursing Taxonomies Marcos Venícios de Oliveira Lopes* Nursing Department, Ceará Federal University, Brazil. L L evel-2 Researcher of the National Council of Technological and Scientific Development, CNPQ, Ceara, Brazil Introduction The use of taxonomies in nursing has been broadly discussed. Its contribution to standardizing the language, implementing information systems, organizing health care practice, and evaluating care effectiveness are often stated in literature as justifications for its implementation in health care services. Particularly in the field of cardiology, few studies have been developed in Brazil to facilitate the incorporation of those taxonomies by clinical nurses. Despite the numerous taxonomy proposals for nursing in Brazil, the most acknowledged taxonomies are those of the North American Nursing Diagnosis Association (NANDA), the Nursing Interventions Classification (NIC) and the Nursing Outcomes Classification (NOC) [1-3]. However, regardless of which classification is adopted, this paper mainly focuses on the need to define the phenomena specific to nursing practice, as a way of making the implementation of clinical studies feasible. Those studies would permit the comparison of data among the various settings in which nursing practice takes place. This paper attempts to include a joint discussion among NANDA, NIC, and NOC taxonomies and the main clinical research designs used in cardiology. The NANDA nursing diagnosis classification is one of the most widespread in the world. In Brazil, notably, it is the phenomenon classification most used by nurses [4]. More recently, the NIC and the NOC have been incorporated into the practice of Latin-American nurses. The NIC was developed by a group of investigators from the University of Iowa (United States) in 4 phases: Creating the Classification ( ); Creating the Taxonomy ( ); Evaluation and clinical enhancement ( ); Use and maintenance (as of 1996). In this classification, nursing intervention is defined as every treatment based on knowledge and clinical judgment performed by a nursing professional to favor the expected patient outcome. Those interventions are subdivided into: a) direct intervention treatment performed through patient interaction; b) indirect intervention treatment without direct patient contact, but for the patient s benefit; c) community intervention directed to promote and maintain population health. Each NIC intervention is composed of nursing activities, defined as specific nursing actions performed to establish an intervention that helps the patient reach the targeted result. These interventions can be found in alphabetical order, by thematic area, NANDA nursing diagnosis and nursing specialty. The classification is composed of 7 domains, 30 classes, 514 interventions and over 12,000 activities. To make it easier to locate interventions and permit an association with NANDA diagnoses, the NIC intervention list is subdivided into: main interventions most likely / most obvious to resolve the diagnosis; suggested interventions associated with the diagnoses, but not very likely for a majority of patients; additional optional interventions apply only to some patients with the diagnosis. The NOC was developed to provide indicators to evaluate a person s health status for problems sensitive to nursing intervention. It was also developed by researchers from the University of Iowa. The NOC defines nursing patient outcome as the state, behavior, or perception of an individual or a community, resulting from nursing interventions, measured during a certain period.

2 The nursing outcomes that compose this classification comprise indicators that describe, more specifically, each state, behavior, or perception of an individual, family or community, and that constitute measures in five-point Likert scales. These scales quantify a patient outcome or the state of the indicator over a certain period, from least to most wanted, and provide a score at a point in time. The evaluation is performed based on the difference between the baseline outcome score and the post-intervention outcome score(s). In terms of structure, the NOC is divided into 7 domains, 31 classes and 330 outcomes, following a conceptual standard similar to that of NANDA and NIC. There are 11 different scales in the NOC, which are exemplified as follows: A Severely compromised Not compromised. Ex.: Cognition B Severe difference from the normal standard No difference from the normal standard. Ex.: Growth F Inadequate Completely adequate. Ex.: Family physical environment G More than 10 None. Ex.: Falls I None Comprehensive. Ex.: Knowledge: diet K Never positive Always positive. Ex.: Self-esteem L Very weak Very intense. Ex.: Health advice. M Never shown Always shown. Ex.: Anger self-management N Severe None Ex.: Anxiety level R Scarce Excellent. Ex.: Social risk control: violence S Not completely satisfied Completely satisfied. Ex.: Quality of life In some outcomes presented in the classification, scale combinations are used to better describe the evaluation of the component indicators: A and N 43 outcomes I and H 6 outcomes M and T 7 outcomes N and A 6 outcomes The importance of these taxonomies goes beyond the description and classification of phenomena, interventions and nursing effectivity indicators. Adopting taxonomies is perhaps one of the few ways of making nursing work comparable, at global level, from the perspective of clinical research. Clinical Research Structure and Implementation Clinical Research is the systematic study that follows scientific methods applicable to human beings, either healthy or ill, in order to characterize a phenomenon, establish risk/cause relationships among phenomena. It also aims to verify effects, safety and tolerance, with a view to establishing the efficacy and safety of an intervention [5]. Independently of any clinical research, some points in its development should be taken into consideration in order to minimize costs and errors, as well as to maximize the benefits. The implementation of any clinical design depends on developing a standard operational procedure (SOP). That is, developing a procedure that aims to permit a process to be performed always in the same way, regardless of the area, so that each stage can be checked. SOPs are common in many health institutions, which use them to describe the implementation stages of a certain technique; for example, changing a surgical dressing. In clinical research, creating an SOP follows a similar pattern. The difference is in its purpose, which is to ensure that the stages of measuring the study phenomenon (either descriptive, longitudinal, or even a clinical trial) are performed in a similar way, thus reducing the possibility of biasing the collected research data. Unfortunately, developing an SOP, though very well structured for implementing the study, is not enough to ensure a research will be successful. Other problems appear when developing clinical studies. Those problems are presented below, along with a few suggestions for solutions:

3 Table 1: Adapted from Hulley et al. These problems and solutions can appear in any research situation and in a range of designs. Two attitudes should prevail in order to guarantee that the study will be successful: having a research advisor and maintaining constant discussion with collaborators. In addition, research groups should not behave as a mere gathering of individuals working isolatedly with distinct objectives and with no common relation. Such a composition would overwhelm the advisor and minimize the chance of developing deep knowledge. It should be noted that the phenomena studied in nursing involve complex human responses, which are difficult to measure, and, in many places, are poorly recorded. Moreover, taking the field of cardiology into consideration, there is one additional problem: the critical aspect of the care required by those patients. In this sense, it is worth emphasizing a point that is considered central in research frameworks and is many times regarded as one of the most difficult stages: the sample plan. This plan requires knowing the variables of interest concerning the target population in the research. This knowledge serves to define the inclusion criteria, which, as a last resort, limit the population to which the inferences developed in the study are applicable. Thus, study results can be considered generalized. Such criteria involve demographic, clinical, geographical, and temporal characteristics. Next, it is necessary to define criteria to minimize influence from external factors. Those factors could influence obtaining reliable comprehensive data and/or including individuals with characteristics that are different from those of the target population. Exclusion criteria comprise individuals with: high probability of loss during follow-up, inability to provide reliable data, high risk of side effects, and a situation that could harm ethical principles. The definition of these points is so important that a simple mistake in these two phases can make it impossible to generalize results and limit their use. At this point, it is important to highlight that the process of selecting the sample individuals should be, preferably, probabilistic. That is, each subject from the population should present the same probability of being selected for the sample. It is sometimes unknown to researchers that most statistical packages develop statistical calculations based on the premise of a simple random sample. When other sampling techniques are used, certain statistics should be obtained differently. SPSS, for example, includes an analysis (block) for complex samples that includes stratified samples. It is common that, while defining the sample plan, the researcher faces various situations that can make it difficult to obtain a simple random sample. A particular case is that of research involving the analysis of nursing diagnoses given to hospitalized individuals. Obtaining a simple random sample requires a rather large number of individuals, which can cause delays to the research or considerably increase costs. In this case, specifically, a consecutive non-probabilistic sample can be used for the sake of convenience. That is, individuals are selected for the sample while they are hospitalized. In order to assure that this strategy reliably represents the target population, seasonal factors must be

4 taken into consideration, since these factors may influence the phenomenon of interest [6]. However, the difficulty in the sample plan does not end by defining the population and the strategies for sample selection and capturing. In fact, establishing sample size raises questions among researchers. Moreover, it is the main reason for many misfortunes, for instance, excessive costs or insufficient allocation of individuals to check for relevant clinical differences. Determining sample size considers the form of measuring the outcome variable(s), the reliability coefficient, the ability to identify significant statistical differences, the magnitude of the effect, the variability of the data, the size of the population, and the accepted random error. Each research design that will be described later in this text presents distinct forms of estimating the sample size. In some cases, the sample size is significantly large, depending on the adopted parameters [7]. Hence, it is suggested that some strategies be used to help minimize sample size: - Using continuous variables - Using paired measures - Using more accurate variables - Using different group sizes - Using a more common outcome These strategies tend to minimize sample size. On the other hand, the researcher should pay close attention to the fact that the sample estimate does not guarantee that statistically significant differences will be identified. A certain clinical view of the data is needed when adopting procedures to minimize sample size. Many times, one certain form of measuring the outcome variable, despite producing a smaller sample, might not be sufficiently adequate to clinically represent the phenomenon of interest. In addition, it is important to avoid calculating sample size considering specific parameters and change the form of measuring the variables during research implementation. In this case, it is possible that the initially calculated sample size is not enough to identify significant differences. It might also produce estimates with a reliability interval that is too large, which makes the practical use of the finding impossible. The next stage in a clinical research includes planning the measurements. In this stage, precision (the ability of a measure to be reproducible) and accuracy (ability to represent what it really should represent) are characteristics that will define the reliability of the data and, thus, study reproducibility [8]. In this study, the researcher s major concern is controlling the biases. To do this, the following strategies can be used: - Standardizing the measuring methods - Training and certifying observers - Optimizing instruments - Atomizing instruments - Repeating the measurements - Performing non-intrusive measures, that is, using forms to measure variables indirectly - Blinding - Calibrating instruments On the other hand, the most important phase of the development is to define the design that will be used. Clearly, this decision is associated with the research question. In short, research designs can be classified in observational and experimental. This paper will be limited to four subtypes: cohort, crosssectional, case control (observational) and clinical trials (experimental). Clinical research designs and the use of taxonomy in nursing Cohort studies Cohort is defined as a group of subjects followed over time to study the behavior of a phenomenon over time, and to verify the respective risk factors for the outcome of interest. It can be developed in a prospective way, in which the subjects characteristics that are believed to influence the outcome are measured and followed periodically. It can also be retrospective, which differs from the first perspective in the sense that measurements were made in the past. Prospective cohorts have the following strengths: a) define incidence and potential causes; b) allow for measuring variables in a comprehensive and accurate way; and c) avoid influence from previous knowledge about the outcome on variable measurement. On the other hand, weaknesses include: a) being expensive and ineffective for rare outcomes; b) requiring more time to conclude data collection; c) excluding persons with a history of the outcome; and d) demanding a great deal of attention to determine if a predictive variable is influenced by the outcome. Retrospective Cohort studies are considered to have the same strengths as prospective studies, and demand more financial resources and less time. However, weaknesses include: a) limited control over the sample and the nature and quality of predictive variables; and b) subject to the bias of the record [7].

5 When planning a cohort study, the following points should be considered: a) determining subjects that are appropriate and available for follow up; b) the sample should have characteristics similar to the population; c) searching for measurement precision and accuracy; d) measuring confounding potentials; e) evaluating outcomes using standardized criteria; and f) following the whole cohort. All these aspects increase the probability of losing subjects during the follow-up. Actually, subject loss is always a significant problem in any longitudinal design. Some strategies can be used to avoid losses, such as: a) excluding subjects with high probability of loss (intend to move or do not intend to return); b) collecting information to locate the subject; and c) keeping periodical contact. Cohort studies are particularly important to learn about the natural development of nursing diagnoses. Following a group that initially does not present a human response can help to clarify supposedly causal relationships and identify important predictors. In this type of design, multiple diagnoses can be studied and their simultaneous occurrence can be analyzed. The association measure, in this case, is relative risk. In addition, other longitudinal analyses can be performed; for instance: time domain analysis and survival analysis. Few nursing studies involving diagnosis from this perspective have been developed. For those interested in further information, two interesting studies involving children with congenital cardiopathy have been published [9-10]. Cross-sectional studies The elementary difference between this kind of study and the previous is that measurements are performed on one single occasion. Therefore, it is useful when the goal is to describe variables and their distribution patterns. This means it is the only design that presents prevalence data. Its strengths include: a) demanding less time; b) having a low cost; and c) having fewer loss risks. Its limitations are also rather significant and include: a) little ability to establish causal relationships; b) few cases to study rare outcomes; and c) limited capacity to establish prognosis or the natural outcome history [6]. Cross-sectional studies are particularly interesting to verify nursing diagnosis profiles. In many situations, the lack of a nursing diagnosis record makes it difficult to adopt taxonomies or even develop care plans structured for specific problems based on clinical evidence. In this case, the starting point is the cross-sectional analysis of nursing diagnoses. Its implementation permits to estimate the percentage of diagnoses and create hypotheses about the relevance and specificity of certain human responses. In addition, in studies about clinical evidence, the analysis of the sensitivity and specificity of the defining characteristics of a nursing diagnosis has shown to be a new perspective in taxonomy research. Cross-sectional studies also permit raising hypotheses about associations between nursing diagnoses. Despite its considerably limited ability to establish causal relationships, the analysis of association strength using statistical coefficients and/or data about chance ratios can help to identify fields for further studies [11 14]. Case-Control Studies Studies with this design are retrospective, based on the definition of one group with and another without the outcome. Its aim is to search for differences in the predictive variables which could explain the occurrence or absence of the outcome. Its strengths include: a) efficiency for rare outcomes; and b) utility in generating hypotheses. Among all observational studies, it presents the largest number of limitations: a) does not estimate incidence, prevalence, or risk; b) only permits studying one outcome at a time; and c) highly susceptible to bias. On the other hand, some strategies can be implemented to minimize those limitations. Regarding sample bias control, the following approaches can be used: a) selecting controls from the same location where cases were selected; b) Pairing even out important factors related to the outcome, but not of interest at that moment. Ex.: gender and age; c) using a population-based sample; and d) using two or more control groups. In terms of differential measure bias control, data recorded before the outcome can be used, then proceeding to blinding [8]. This kind of design is still rarely used in taxonomy studies. It is also important to define clinical differences among groups with and without a nursing diagnosis. Unfortunately, the fact that it depends on a previous record makes it unfeasible for most research proposals with this design. Until date, it has been emphasized that diagnosis taxonomy benefits most from observational research design. In fact, the diagnosis phase is treated like this because it is clearly an identification phase. That is, until now, there is no work involving intervention or intervention evaluation (results). On the other hand, the NOC proposes to establish baseline levels. Actually, these levels measure the intensity of an initially identified human response. This topic can be addressed with the above described research designs. It is different because of its need to identify the presence of nursing diagnoses, establish adequate scale and NOC indicators, as well as to create the operational indicators

6 in a reliable way. This is another approach that helps to solve doubts about the clinical importance of a certain nursing phenomenon. Clinical Trials Unquestionably, the gold standard in clinical evidence is established by clinical trials. The advantage of this type of design is it capacity to demonstrate causality. However, there are disadvantages, which include: high cost, time consumption, the fact that it addresses a restricted clinical issue and the potential of harming the patient [15]. The difficulties to implement a clinical trial begin when selecting the patients. In this stage, one should consider: the primary outcome rate; the expected efficacy of the active treatment; the capacity of generalizing the findings; the facility of recruiting subjects; and the probability of compliance to treatment and follow-up. The following are some of the reasons to exclude subjects from a clinical trial: - One of the treatments may be harmful (Ex. Unacceptable risk of allocating a placebo) - The active treatment is not effective (Ex. There is no risk for the outcome) - Low probability of compliance with intervention (Ex. Low compliance in the pretreatment period) - Low probability of completing the follow-up period (Ex. Plans to move) - Practical problems for participation (Ex. Mental disease) Next, the baseline variables need to be measured, that is, to characterize patients in terms of the clinical variables of interest at the start of the study, in order to compare those measures after applying the interventions. In this stage, the following should be accomplished: a) Collect information for posterior subject location; b) Describe patients (preserve homogeneity at baseline); c) Measure variables that represent risk factors for the outcome or which could define subgroups; and d) Measure the outcome variable [16]. In order to compare the intervention and control groups, a randomization technique must be used. This procedure should allocate treatments randomly in an inviolable way. Preferably, an equal number of subjects would be allocated to each treatment, and a table can be used with random numbers, computational algorithm, or sealed envelopes. Among special randomization techniques, the following are highlighted: a) Randomization in blocks, in which the randomization is performed using blocks with previously determined sizes, to ensure an equal distribution between study groups; and b) Randomization in blocks and stratified, which aims to distribute an important outcome predictor in a more even way [6]. Another important aspect of clinical trials is the application of interventions. The risk of bias must be eliminated in order to ensure result generalization. In this sense, blinding is one of the most important aspects, since it eliminates the influence from confounding variables present at the moment of randomization (Ex. The researcher might pay more attention to the group receiving the active treatment) and avoids bias in outcome assessment (Ex. The researcher might tend to find outcomes in the control group). Blinding difficulties include logistics (finding comparable treatments with similar appearance), and avoiding that patients and the team are able to tell the difference between the effects of the designated treatments. It is interesting to invite patients, at the end of the study, to try to guess the treatment designations. When choosing the intervention, the following should be considered: the balance between effectiveness and safety; the feasibility of blinding; the choice between one intervention or a combination of interventions; the capacity of incorporating the intervention in everyday practice; the choice of the control. The ideal process would be to compare without receiving the active treatment and with blinding. Since this is impossible, the standard treatment can be administered to all participants, or a group following the standard treatment can be compared to another alternative group (equivalence trial). In the second case, a larger sample is required. Follow-up and evaluation of compliance to the protocol is one of the most important activities for a successful clinical trial. Therefore, it is important to choose the subjects with stronger changes in compliance, facilitate intervention, make consultations convenient, avoid measurements that cause pain or discomfort, encourage subjects compliance with the protocol, and locate subjects lost during follow-up. The final variable to be considered in this type of study is outcome measurement. There are often doubts about adopting a direct clinical outcome (for example the identification of a nursing diagnosis) or a substitute outcome (clinical indicators of an NOC results scale). This choice should be made according to the measurement reliability and facility offered by the outcome variable. NOC scales have the advantage of characterizing the outcome in multiple aspects under five scale levels. On the other hand, directly using a nursing diagnosis makes interpretation easier, but requires the creation of operational definitions for the defining characteristics.

7 In addition, the outcome variable statistic characteristics can facilitate posterior data analysis and conducting the trial (continuous variables maximize the strength of a study with smaller sample size). The number of outcome variables is also a reason of doubt for many researchers. The authors of the present study support the opinion that the number of outcome variables should be the smallest possible. This is important because it defines the size of the sample in which the trial will be evaluated. Usually, the sample calculation is performed with all outcome variables, and the largest sample size calculated is chosen. One must not forget to confirm the self-reported outcomes and include the outcomes associated with side effects. The above description denotes the complexity of implementing a clinical trial. This research design concentrates a major part of current efforts in nursing development. It is also in this design that taxonomies have the strongest chance of helping nursing. The three basic nursing process stages are present in clinical trials: identifying the diagnosis; implementing intervention; and evaluating the result. Thus, elaborating a research protocol at this level requires the construction of a standard operational procedure that describes: a) patient allocation, with an operational definition of the defining characteristics and its definition process for the presence of a nursing diagnosis (some articles discuss the logic involved in this stage [17]); b) the construction of an intervention protocol, as well as its form of implementation in order to assure application in an even way; and c) the definition of the NOC scale(s) to be used, including clinical indicators, operational definitions, period between evaluations and types of measurement. Final Considerations Nurses around the world have not reached a consensus about the use of nursing taxonomies. Although, in the present study, the authors have described the importance of using taxonomies in clinical studies, and only three were described (NANDA, NIC, NOC), it does not mean that those are the only or best ones. What is important is to assure that the indicators used are reliable, reproducible and common to one s practice. The comparability of the presented strategies can be incremented by adopting a common language that would favor communication between nurses. On the other hand, the specificity of many regions in the world, the difficulty of records, and the lack of familiarity with clinical research methods are factors that could make that adoption difficult. Yet, the authors of the present study believe that the advance of evidence based nursing requires the definition of a language that allows for comparisons among phenomena. Thus, the authors would like to state their expectation of a reflection process about the position towards this important issue of nursing knowledge. Moreover, they present an invitation to develop collaborative research. The term person is used in this manuscript with the meaning of individual, family, community or any other type of group consistent with the population attended by nursing professionals. Bibliography 1. North American Nursing Diagnosis Association (NANDA). Diagnósticos de enfermagem da NANDA. Porto Alegre: Artmed, Mccloskey J C, Bulechek G M. Classificação das intervenções de enfermagem. Porto Alegre: Artmed, Johnson M, Maas M, Moorhead S. Classificação dos resultados de enfermagem. Porto Alegre: Artmed, Silva VM, Oliveira TC, Damasceno MMC, Araújo TL. Linguagens da sistematização da assistência de enfermagem nas dissertações e teses dos catálogos do Centro de Estudo e Pesquisa em Enfermagem. Online Braz J Nurs 2006; 5(2). Disponible en: 5. Lousana G. Pesquisa clínica no Brasil. Rio de Janeiro: Revinter; Hulley SB, Cummings SR, Browner WS, Grady D, Hearst N, Newman TB. Delineando a pesquisa clínica: uma abordagem epidemiológica. Porto Alegre: Artmed, Jekel JF, Elmore JG, Katz DL. Epidemiologia, bioestatística e medicina preventiva. Porto Alegre: Artmed, Greenberg RS, Daniels SR, Flanders WD, Eley JW, Boring JR. Epidemiologia clinica. Porto Alegre: Artmed; Silva VM, Araújo TL, Lopes MVO. Evolution of nursing diagnoses for children with congenital heart disease. Rev Latino-Am Enferm. 2006; 14(4): Silva VM. Caracterização de diagnósticos de enfermagem em crianças com cardiopatia congênita: estudo num hospital especializado em doenças cardiopulmonares. [Disertación]. Universidad Federal de Ceará Silva VM, Lopes MVO, Araujo TL. Diagnósticos de enfermería y problemas colaboradores en niños con cardiopatías congénitas. Rev Mex Enferm Cardiol. 2004; 12(2): Silva V M, Lopes M V O, Araujo T L. Asociación entre diagnósticos de enfermería en niños con cardiopatías congénitas. Enferm Cardiol. 2004; 11 (32-33): Silva VM, Lopes MVO, Araujo LT. Análisis del diagnóstico enfermero. Patrón respiratorio ineficaz en niños con cardiopatías congénitas. Enferm Cardiol 2006; Año XIII (38): Paiva GS, Lopes MVO. Respuestas humanas identificadas en pacientes con infarto agudo de miocardio ingresados en una unidad de terapia intensiva. Enferm Cardiol 2005; Año XII (36): 22-

8 Pocock SJ. Clinical trials: a practical approach. Chinchester: John Wiley & sons, Vieira S, Hossne WS. Metodologia científica para a área de saúde. Campus: Rio de Janeiro Lopes MVO, Silva VM, Araújo TL. Desenvolvimento lógico-matemático do software ND. Rev Latino-am Enferm 2004; 12(1): CV of the author - Doctor en Enfermería - Post-Doctorado en el Departamento de Estadística e Investigación Operativa de la Universidad de Valencia - Profesor Adjunto del Departamento de Enfermería de la Universidad Federal de Ceará, Brasil - Investigador 2 del Consejo nacional de Desarrollo Científico y Tecnológico CNPQ, Brasil Publication: October 2007 Your questions, contributions and commentaries will be answered by the lecturer or experts on the subject in the Cardiovascular Nursing list. Please fill in de form and Press the "Send" button. Question, contribution or commentary: Name and Surname: Country: Argentina address: Re-type address: Send Erase CETIFAC - Bioingeniería UNER - Webmaster - HonCode - pwmc Updated: 10/01/2007