With respect to the McGuckin studies neither randomization nor sample selection is ever discussed. In fact sampling per se is not presented except for cursory mention in the results section. Again, without proper identification as to the sampling method implemented, the reader is ever cautious as to how legitimate the results will be. Not wanting to pre-empt the discussion on statistical methodology, mention must be made at this time with respect to the Poisson Regression statistical tool selected for use in these two studies in terms of sampling. This particular regression technique, if utilized properly, requires the sample size to be determined on the basis of the square root transformation of the Poisson random variables. More specifically, the formula for calculating the sample size of the Poisson variables is as follows:
The data received from this calculation will give the research investigator the number of sampling units per group that are needed. Unfortunately however the reader is not presented information with respect to the Poisson random variables selected not the unit sample number per group.
Statistical Tool and Selected Variables. Both studies made use of the Poisson regression technique to ascertain their results. When a statistical tool is chosen the tool must conform to the theoretical requirements of the selected process. According to the authors the Poisson regression method was used to test the significance of the intervention. Actually, researchers to test for the interaction of selected variables use a Poisson model and not to test for differences between and amongst a study's selected independent and dependent variables. In essence the Poisson regression model permits predictability of the results. At no time during either research project did the authors state, or allude to, the need to establish the predictability of the Partners in Your Care assessment instrument. The appropriate conclusion drawn, therefore, is that a more user-friendly statistical tool should have been employed - a modified orthogonal analysis of variances wherein nested variables are also considered.
As mentioned earlier, all research investigations must clearly define the variables under investigation. In descriptive research studies this involves the identification of the independent and dependent variables. Although both studies have independent (treatment) and dependent (change) variables they are never clearly identifies. Identification must take place prior to the establishment of the selected statistical tool. Further, when discussing or presenting the independent and dependent variables the research investigator is obligated to inform the reader as to whether or not differences, relationships or effects are being sought. It is not until the data analysis section is the reader advised that the study is seeking to determine intervention effects of the Partners in Your Care model. Again, the fact that the authors make use of the work 'intervention' is indicative of determining effects - which the Poisson regression model is not equipped to handle. Had the research investigators wanted to seek the predictability of the aforementioned model then consideration for using the Poisson model can be established. Furthermore, as the variance of the Poisson random variable is equal to the mean of the response recorded (i.e., response to the Partners in Your Care assessment instrument) and must be reported by the investigators as the regression model is set up to meet normality and variance homogeneity assumptions. Again, no such presentation was made in either study.
Selected Measurement Instruments. When employing any type of measurement instrument the instrument must be evaluated in terms of reliability and validity (Stanley & Hopkins, 1972). Reliability comes from the necessity for dependability in measurement. If the data is not dependable, any conclusions drawn will lack soundness. In order to achieve reliability the research investigator must maintain the integrity of the testing by ensuring that the same results from a testing instrument will deliver the same set of results when tested by a comparable instrument (Ohlson, 1998). As there exists no reported studies confirming the reliability of the Partners in Your Care instrument, the acceptance of the results is suspect. A measurement instrument that has not been properly serviced or calibrated will yield biased or erroneous data and faulty conclusions will be drawn. The error commonly associated with the measurement instrument is commonly called systematic variance or error (random) variance. In reference to the two studies conducted by McGuckin, et al. there is no reference to how the assessment instrument achieved the necessary required reliability. The reader is simply expected to accept the measurement instrument as a reliable indicator of the variables being examined.
In terms of instrument validity the authors are first obligated to define the variables under investigation followed by a presentation of the measurement instruments validity. The types of measurement validity all researchers must be aware of are: content, predictive, and construct validity. Content validity is guided by the question: Is the substance, content, or design of the measure representative of the subject matter being measured. Simply put; is the instrument measuring or testing what it should be testing; and if so to what degree of success. As no information was presented by the research investigators with respect to the content of the instrument, except for a listing of eight questions, the reader is unable to ascertain if the instrument is measuring what is intended. Predictive validity refers to the effect and independent variable (patient request to wash hands) has on the dependent variable (infectious diseases in an inpatient ward). As there exists not information as to previous studies being conducted by McGuckin, et al. The acceptance of the instrument having predictive validity is circumspect.
Construct validity, unlike both content and predictive validity, is necessary in determining the relative variance contained within the testing instrument. Construct validity actually is representative of the influences of the different degrees or strengths (nested variables or outside influences) of the instrument. For example, the question might be well of one asking whether or not English speaking vs. ESL participants presented the eight question differently to nurses and/or physician. In the end is to establish an instrument that is error free. Unfortunately the measurement instrument utilized by the research investigators does not present, nor reflect upon, the validity coefficients of the Partners in Your Care testing devise.
Research Summary. Both the McGuckin, et al. studies conclude (even though is no section labeled as such) that an important way in which to reduce the spread of infectious disease in hospitals is to involve patients in the process of either reminding the healthcare giver to wash their hands before being examined or at least to ask the healthcare giver to wash their hands after being examined. Although the program empowers patients to be proactive in their own healthcare program, provides a continuous means for education of healthcare providers, and does not increase hospital expenditure the study did not provide substantive evidence with respect to what infectious diseases were most affected through utilization of the Partners in Your Care program.
Evaluation Summary. For the purpose of brevity and comprehensiveness the following points are reflective of the value of the two articles reviewed:
Neither article stated in acceptable form a research question or testable null hypothesis.
Neither article identified the independent or dependent variables.
Neither article defined the terms important to an understanding of what was being investigated.
Neither article presented a rationale for the statistical tool used to analyze the measurement data.
Neither article applied the necessary requirements for sampling in conjunction with the chosen statistical tool.
Neither article explained the reliability and validity concepts of the used measurement instrument.
Neither article clearly neither identified the relationship between patient involvement and infectious disease control nor was they're any identification of what infectious diseases were controlled.
The articles were replications but considered separate and as such this reviewer firmly believes that the only reason for conducting the study in America in 2001 was solely for the purpose of receiving another journal entry.
Armitage P. And Berry, G. (1994) Statistical Methods in Medical Research (3 rd edition).
London: Blackwell Publishing.
Connor, E.F., Hosfield, E., Meeter, D. And X. Nui. (1997). Tests for aggregation and size-based sample-unit selection when sample units vary in size. Ecology 78: 1238 -1249.
Corning, S.P. (2002). Profiling and developing nursing leaders. Journal of Nursing
Administration, 32, 373-375.
Ferguson, George a. (1964). Statistical Analysis in Psychology and Education. New York:
McGraw-Hill Book Company.
Kerlinger, Fred (1964). Foundations of Behavioral Research. New York: Holt, Rinehart and Winston.
McGuckin, Maryanne, et al. (1999). Patient education model for increasing hand washing compliance, American Journal of Infectious Control; 27:309-314.
____ (2001). Evaluation of patient-empowering hand washing hygiene programme in the UK,