Basic Fundaments of Designing a Quality Research

Introduction/ Objectives: There exists a large number of papers that are published in around 500 journals within dentistry. Most of these published papers are observational studies with fewer experimental studies. Main objective of the present article is to provide basic knowledge of the study designs, qualitative and quantitative data, framing questions for designing a research for different types of studies, sampling methods, bias that occur during conducting the research. Sources/data: We made a survey of two renowned orthodontic journals (American Journal of Orthodontics and Dentofacial Orthopedics, Angle of orthodontics) from January 2009 to October 2014 and found that most of the studies were observational (65.3%) and the rest were experimental studies (34.6%). Selection criteria: The original research articles from these journals were selected and categorized into observational and experimental studies. Case reports, clinician’s corners etc. have been excluded. Conclusion: In this article, information about the step-wise approach is provided regarding research methodology relevant to our clinical research. Key wordsResearch, Study design, validity, reliability, random error, systemic error, bias. Jain S et al. Basic Fundaments of Designing A Quality Research. 89 Journal of Advanced Medical and Dental Sciences Research |Vol. 3|Issue 1| January-March 2015 To understand the basis of research methodology, one must understand following terminologiesRandom errorError is the difference in true value and study value. Random error describes the role of chance. Random error creeps in because of Sample variability (as a subset of the population is sampled sampling error), Subject to subject variation and measurement error leading to inconsistent value on repeating the experiment. They can be either positive or negative but average is not affected. Random error can be reduced by taking large sample size, repeating the experiment and averaging. (Figure-1) PRECISIONPrecision is opposite of random error, if random error is less, more precise is the study. Precision is a measure of consistency and is the function of random error and confidence required. Lesser the type-I error more precise is the study. SYSTEMIC ERROR (BIAS) Errors are not alone the consequence of chance alone, but occur due to inappropriate patient selection method/criteria. This can be controlled by a good study design and conduct of the experiment. (Figure-2) Validity is opposite of systemic error. Lesser the systemic error, more valid is the study. Systemic error tends to be consistently either positive or negative and thus affect the average. VALIDITY AND RELIABILITY – A Study should be valid i.e. it should measure what it claims to measure and reliable that is it should measure same readings even if measured on two separate occasions. A method is valid (accurate) if we are able to control systemic bias. For e.g.If an index is used to measure dental caries, it must only measure the caries and not enamel hypoplasia (validity). The index should give the same score if the patient is examined by two different examiners (interexaminer reliability) or by the same examiner on two different occasions (intraexaminer reliability). TYPE 1 ERROR (α ERROR)Type 1 error is observing a difference when there is actually no difference (false positive results). The null hypothesis is rejected erroneously. P VALUEIt is the probability of committing Type – 1 error . Type -1 error is fixed in advance. The upper limit of tolerance of Type -1 error is known as the level of significance α. Usually the probability (α) is set either at 0.05(5%) or at 0.01(1%).The Lower the level, the larger is the sample size required. For critical results, the study should be more precise hence α – error is set at a lower level. If the p-value is less than 0 .05, it means that the null hypothesis is rejected. If p-value is 0.01 it means that the difference is statistically significant and there is 1% chance that the difference is by chance. POWER(1-β) OF STUDYPower is probability that a test will correctly identify the difference if it is there. Usually most study accepte the power of 80% i.e. 20% chance of missing the real difference. Sometimes a larger study power is set at 90% i.e. 10% possibility of false negative results –β error. Type II or β error is missing the real difference when it is there. Power analysis should be conducted to determine the appropriate sample size. The larger the sample size, the study will have greater power to detect the significance of difference or effect of association. Power proportionally increases as the sample size for study increases. RANDOMIZATIONIt is a statistical procedure by which participants are allocated into groups(study and control group) to receive or not to receive an intervention.It is a method for allocating subjects to different groups in an unbiased manner. (See Table-1) Each individual gets equal chance to be selected in either group. Randomization is an attempt to eliminate bias and allow for comparability. Jain S et al. Basic Fundaments of Designing A Quality Research. 90 Journal of Advanced Medical and Dental Sciences Research |Vol. 3|Issue 1| January-March 2015 BLINDINGIt is a basic tool to prevent conscious as well subconscious bias. In single blinding study subjects do not know the type of treatment they are receiving while in double blinding subject and investigator both do not know about the type of treatment. In triple blinding subject, investigator and analyzer all are not aware of type of treatment subjects are receiving. (See Table-1) SAMPLING Sampling can be probability or nonprobability type. In probability type every individual has an equal chance of getting selected in the study. Probability sampling methods are – simple random sampling, stratified random sampling, multistage sampling and systematic sampling. (See Table-2) While in non-probability sampling every individual is not ensured an equal chance of getting selected. FACTOR AFFECTING SAMPLE SIZE CALCULATIONsampling size determination is one of the most essential part of any research/study. Sample size calculation provides the appropriate no. of participants to be included in the study.There are many factors which affect the sample size, some of them are shown in Table-3. TYPES OF STUDIES DESCRIPTIVE STUDY – Descriptive studies are the first phase of any epidemiological studies focused on entire population. These studies describe the pattern of occurrence of disease. They provide data with regards to the type of disease problems and their magnitude in the particular community, aetiology of disease, planning preventive and curative series. Descriptive studies are used to formulate hypothesis. CROSSSECTIONAL STUDIES – Cross-section studies are used to find out the prevalence of disease. e.g. Prevalence of Cleft Lip and Cleft Palate in India. Crosssectional study design is best to be used for diagnosis/ screening, occurrence, surveys or establishing norms. Surveys and cross-sectional studies are planned to observe various pattern like quality of life. Subjects are merely observed and no attempt is made to impose treatment on them. Cross-sectional studies are descriptive studies which can suggest hypotheses for future cohort and case control studies. ANALYTICAL STUDIES – Analytical studies are used for testing the hypothesis. Analytical studies, ascertain statistical association between two things by the test of significance. Analytical studies can be either observational type or experimental type. In observational type, we simply observe either retrospectively (case control study) or prospective study (cohort study). So study design is either backward looking or forward looking. If our research question is What is the role of flossing teeth in dental care? Then we can design observation study which can be prospective or retrospective. In a prospective study we include the patients who will be flossing the teeth regularly and those not flossing the teeth and examine their dental hygiene at regular time interval. While in retrospective study we include records of subjects who were already flossing teeth regularly and those who are not flossing, we examine them and find the dental carries record. EXPERIMENTAL STUDIES These are prospective studies. An experimental study aims to control all relevant variables while altering only the variable under investigation. We test the effectiveness of a drug or a treatment with the conventional methods. The experimental studies are either randomised (avoiding systemic bias) or nonrandomised. If our research question is – Which filling material is best in the primary dentition? Are glass ionomer and resin-based fissure sealants – equally effective? To answer these questions, we can design experimental study including two groups of Jain S et al. Basic Fundaments of Designing A Quality Research. 91 Journal of Advanced Medical and Dental Sciences Research |Vol. 3|Issue 1| January-March 2015 patients using GIC in one group while in the other group, resin based fissure sealant is used and in both the groups, we may find the effectiveness of these materials in prevention of carries (caries). To design a study, we take different groups of primary dentition and use different filling materials and in each group, we can compare their effectiveness in terms of reduction of caries. Graph-1 shows basic classification for the types of research studies. CASE-CONTROL STUDIES (RETROSPECTIVE) – First approach to test a hypothesis is a case control study. They do not establish the etiological factor but simply find the association between the two by calculating the odds ratio (an estimate of relative risk). They are suitable for examining the rare disease. E.g. risk factor associated with root resorption in the patients who have undergone orthodontic treatment. They are not used for calculating the incidence of disease because exposure and risk have already occurred. So we do not know the population at risk. If odds ratio is 8:1, risk of outcome (eg.malocclusion) is 8 times greater for subjects with any risk factor (eg.abnormal habit). OR < 1 Beneficial exposure OR> 1 Hazardous exposure OR= 1 No association (no difference in exposure between case & control) COHORT STUDYCohort is a group of individuals sharing common characteristics (study cohort and control cohort). These groups are identified prior to the appearance of disease. These observational studies are used to explore the effect of exposure on outcome(mostly prospective study). For e.g. Natural variation in exposure and intervention are investigated. Long cohort studies(mostly prospective) are well suited for measuring incidence of disease. Cohort study is used to find the strength of association of cause and effect by assessing risk ratios and attributable risk from the 2x2 contingency table. (See table4)