Data Collection Plan
To successfully measure the efficiency of the intervention, it is necessary to determine the information that is directly relevant to the identified goals and objectives. In this particular case, 30-day readmissions are one of the reliable indicators of the hospital’s discharge process’ success rate. Therefore, the information can be defined as data on the admission of the patients previously discharged from the hospital in question and admitted to the respective hospital with a complaint that corresponds to the condition stated as an initial health issue. The scope includes data on both planned and unplanned admissions and uses an inclusion criterion of the index admission as a baseline.
However, it should be stressed that the principal diagnosis should not be used as one of the inclusion criteria since it is irrelevant in terms of the quality of the discharge process. In simple terms, the readmission contains risks for the patient and creates adverse effects for the healthcare provider with no regard for the cause of the event. Therefore, such an approach would allow for a stronger emphasis on patient-centered care and at the same time ensure an overall increase in quality of care within the hospital.
Considering the nature of the identified information, it would be reasonable to choose medical records as the main source of data. Medical records contain relevant and definitive information that can be quantified and analyzed to produce highly reliable results. The data in question should be collected in three tiers. First, the inpatient data can be retrieved. This set includes the information on the diagnoses, dates of service, diagnosis-related groups, procedures with respective ICD codes, the information about the care provider, and the demographic information of the patients. Second, the outpatient data is necessary to collect.
This set would include a range of non-inpatient services such as emergency room care, diagnostic procedures, and surgical interventions, as well as services conducted in the ambulatory diagnostic centers and outpatient department of the hospital. Third, the records of the physician services performed face-to-face with the patient (CMS, 2016). As such, this source would not include laboratory tests and ambulatory services. Besides, the third source would disregard the setting of the provided service. Such disaggregation would allow for a comprehensive nature of the retrieved data and ensure the inclusion of all relevant aspects of the readmission process.
The source of the data suggests that all patients within a certain timeframe would be subject to sampling procedure as long as they meet the inclusion criteria of being readmitted to the hospital within the 30-day timeframe. Therefore, as long as the project is confined to a single hospital, the size of the sample is determined by the amount of the medical records pertinent to the issue. However, it should be mentioned that such sampling size choice is subject to several limitations.
First, the size of the sample depends primarily on the number of occurring readmissions within a single organization, which, in turn, opens up the possibility of inconsistencies with the average rates. While this limitation can be addressed by measuring the relative readmission rate and comparing it to the respective results of other organizations, such adjustment is outside the scope of this project and should be performed separately. Second, the actual size of the sample depends directly on the size of the hospital and the demographics of the local population, which may or may not be representative of the total population of the state or a country.
In other words, the obtained data and the conclusions on the success of the intervention cannot be reliably used until the sample is established as representative of the audience of interest. The principal characteristics of the sample include the age (younger than 65 years), length of stay (the exclusion of same-day readmissions due to incompatibility), the absence of transfers to another facility, the occurrence of in-hospital death during the initial admissions, and the additional admissions (due to their differences from readmissions).
Based on the identified criteria and characteristics, the following methodology is suggested. The data will be collected from the hospital databases that contain information on the hospitalization of the patients from 2016 and 2017. The hospitalization data is expected to include the information on the index admission of the patients, their eventual discharge, and a medical record of 30 days following the event. The retrieved set is further refined by excluding the individual entries which do not comply with the stated exclusion criteria such as age and same-day readmissions. The cases where the principal and readmission diagnoses do not match are not excluded to further stress the quality improvement side of the project.
The sample is then further refined by identifying the cases where additional admissions were part of the outcome and thus cannot qualify as a gap in the quality of the discharge process. Finally, the data is verified through the comparison with the administrative data available on the selected patients. If the administrative data produces a reliable estimation of the results derived from the medical records data, the data is considered valid and can be used for analysis. In this way, the compliance of the discharge tool with the standards of clinical research can be established.
To meet the deadlines of the project, the entire data collection process is planned for six weeks. The location of the necessary records is expected to last for one week, followed by the verification process and the location of the administrative data, with the final week reserved for the establishment of the discharge tool compliance and the confirmation of the initial calculations.
Data Analysis Plan
Once the necessary data is retrieved and validated, it is possible to proceed with the analysis procedures. Quantitative analysis has been chosen to perform the analysis since it yields more reliable results and demonstrates greater efficiency when applied to the large data samples (e.g. the cohort within a single hospital). The data analysis procedure is expected to illustrate the readmission rate within the hospital which, in the context of the study, can be viewed as an intercept specific to the hospital in question as well as a function of the organization’s characteristics and the demographics observed at a patients’ level.
In this way, the procedure would acknowledge the outcomes observed at the hospital level and determine the correlation between them (McIlvennan, Eapen, & Allen, 2015). Also, it would allow substantiating the modeled suggestion that the difference in the outcomes within a specific healthcare facility can be used to evaluate the systematic improvements in outcomes on a larger scale. To ensure the applicability of findings, risk adjustments need to be introduced into the procedure. Specifically, the inpatient and outpatient data will be used in the analysis, backed by the face-to-face encounters with the physicians and a range of secondary diagnoses.
The multi-faceted nature of the dataset would allow for the calculation of the risk-adjusted covariates. The ICM codes would then be grouped into mutually exclusive categories, which, in turn, can be utilized for the formulation of the condition categories and, by extension, the introduction of the variables for risk adjustment. Once these variables are available, it would be possible to estimate the risk factors and determine the presence of random effect within the facility.
Using the model above, it would be reasonable to evaluate the possibility of readmission for any given individual patient within 30 days. After this, it becomes possible to calculate the average probability of the readmission by summing up the results for all patients within the hospital and dividing the result by the summed probabilities of each specific condition category. The obtained result would illustrate the average readmission in a scenario where any given patient would be exposed to the risk of readmission with a random diagnosis. Finally, based on the received result it would be possible to calculate a standardized readmission rate by multiplying the obtained ratio by the overall rate available from the data. Such a procedure would mitigate the limitations of the possibly inadequate sample by scaling the findings to the degree where they would apply to both smaller and bigger hospitals.
The primary data analysis tool used in the project is a hierarchical regression model. The rationale for this choice is the suitability of the tool for the chosen data collection methods and the expected reliability of the results. Specifically, the regression coefficients will be estimated for each patient in the dataset, multiplied by the characteristics of the readmission, and added to the in-hospital intercepts. In this way, quantity can be converted to probability and, after summing the respective outcomes of all patients, the average probability can be derived (Mull et al., 2013).
With this information on hand, it would be possible to model the predicted number of readmissions by summing the individual predicted readmission rates of all patients. According to the hierarchical model, the individual predicted readmission rate could be calculated by adding the specific condition categories to the estimated regression coefficients for each participant about the observed characteristics. The described tool would summarize the observations within a single hospital. The risk factors of the hospital-specific random effect would have to be introduced to exclude the distortion of results due to the clustering of observations.
To summarize the obtained data in the accessible format, the run chart will be used as an additional tool. The run chart has several benefits that make it highly suitable for the project. First, it summarizes the progress of a project in a visually comprehensible manner and allows for a flexible adjustment depending on the frequency of the data retrieval. Next, it provides the means of identifying the onset and termination of a cycle or a time-specific trend during the project’s duration. Finally, given sufficient time and frequency of the inquiry, it can reveal a non-random pattern in the project’s development, which can be used for the elimination of undesirable effects or the identification of the overlooked benefits (Houser, 2016).
Considering the stated goal of decreasing the hospital readmissions by 50% in the six months of the project duration, it is possible to suggest the following run chart as an example of the project’s outcomes.
The x-axis, in this case, corresponds to the duration of the project and is marked by month. The y-axis identifies the average discharge rate. For exemplary purposes, the initial value is set at 19%, which transpires to the target line of 9.5%. The data in the example is retrieved every month.
Action Plan
To reach the desired level of quality, a specific methodology is recommended for implementation. Considering the complexity of the issue and the number of factors involved, a cyclical method such as plan-do-study-act (PDSA) is deemed the most viable option. Such a method allows for a systematic improvement due to the seamless nature of the process. Also, its effectiveness is widely recognized in the field of healthcare. PDSA has been successfully used in similar projects in the past with the documented positive results (Poston, Dumas, & Edlund, 2014). For the current project, its use can be outlined by its principal components.
First, the goals and objectives of the project must be laid out clearly and concisely, with quantifiable milestones and reasonable deadlines for each stage. Each area of improvement should be addressed by a specific evidence-based practice followed by the rationale for its choice and the estimates of the expected benefits. Importantly, the data collection measurements also need to be identified at this stage, with specific dates for each inquiry.
Second, the planned changes are set in motion by a designated team. Simultaneously, the observation of the immediate outcomes and compliance with the deadlines is to be performed. The former can provide an additional level of control of the process as well as a possibility of readjusting minor details in the case when they are incompatible with the expectations. The latter ensures the consistency of the project’s implementation and, by extension, the reliability of the outcomes.
Third, the obtained results are analyzed and laid out in an accessible manner for the participants to detect the causes for shortcomings if the objectives are not met.
Fourth, based on the conclusions derived from the observations, adjustments are developed for implementation in the following cycle. In the scenario where all objectives are successfully met, this stage can include the development of a more effective version of the intervention or, in some cases, the decision to use the unaltered version.
The main criterion of the project’s success is the decreased patient readmission rate. Therefore, the assessment and measurement of progress both during and after the intervention can be based primarily on this metric. The process for assessment will include the retrieval of data from the medical records, the verification of data through the administrative sources, and the analysis using a hierarchical regression model. Once the results are available, they will be laid out in the form of a run chart. In this way, the progress achieved by the project as well as the dynamics of the process can be assessed. The evaluation process should then be repeated in three and six months following the intervention to determine the long-term effectiveness of the intervention. Such an approach would improve the understanding of the procedure’s effects and allow for more precise modeling before its implementation in similar settings.
References
CMS. (2016). Readmissions reduction program (HRRP). Web.
Houser, J. (2016). Nursing research: Reading, using and creating evidence (4th ed.). Burlington, MA: Jones & Bartlett Learning. Web.
McIlvennan, C. K., Eapen, Z. J., & Allen, L. A. (2015). Hospital readmissions reduction program. Circulation, 131(20), 1796-1803. Web.
Mull, H. J., Chen, Q., O’Brien, W. J., Shwartz, M., Borzecki, A. M., Hanchate, A., & Rosen, A. K. (2013). Comparing two methods of assessing 30-day readmissions: What is the impact on hospital profiling in the veterans health administration? Medical Care, 51(7), 589-596. Web.
Poston, K. M., Dumas, B. P., & Edlund, B. J. (2014). Outcomes of a quality improvement project implementing stroke discharge advocacy to reduce 30-day readmission rates. Journal of Nursing Care Quality, 29(3), 237-244. Web.