Avoidable Readmissions For Patients Hospitalized With Community-Acquired Pneumonia

Overview

From 10% to 30% of patients hospitalized with community-acquired pneumonia (CAP) are readmitted within 30 days of discharge. These readmissions have negative consequences for the patients and the hospitals where they are treated, including impaired quality of life, exposure to hospital-related adverse events, and increased resource utilization. Risk-adjusted readmission rates can be easily computed and tracked from computerized hospital discharge data, using validated models. As part of the Hospital Readmission Reduction Program (HRRP) effective in fiscal year 2013, United States hospitals with higher than expected 30-day readmission rates after pneumonia hospitalization have been subject to financial penalties from the Center for Medicare and Medicaid Services (CMS). The underlying logic of the HRRP is based upon the notion that short-term readmission is often a preventable adverse outcome, reflecting suboptimal quality of care during index hospitalization. Yet, published evidence suggests that less than one in four all-cause readmissions are deemed avoidable. Because only avoidable readmissions can be influenced by interventions designed to decrease readmission rates, avoidable readmission is a more relevant metric than all-cause readmission for tracking quality of hospital care for pneumonia. The purpose of this study is to develop an administrative data-based risk prediction model for identifying potentially avoidable readmissions within 30 days of discharge for patients hospitalized with CAP. R3P is a retrospective observational cohort study of consecutive adult patients discharged from two hospitals with a diagnosis code of CAP. Data sources include routinely collected hospital discharge data and retrospective chart reviews.

I. Objectives The purpose of this study is to develop an administrative data-based risk prediction model for identifying potentially avoidable 30-day readmissions of patients hospitalized with community-acquired pneumonia (CAP). The focus is on potentially avoidable readmissions because not all hospital readmissions can be influenced by interventions designed to decrease them. The specific aims of this project are: 1. to assess the accuracy of International Classification of Disease, Tenth Revision (ICD-10) discharge diagnosis codes for CAP using a structured retrospective chart review as the reference method. 2. to estimate the rate of all-cause readmissions in the same hospital within 30 days and one year of discharge for patients hospitalized with CAP. 3. to estimate the percentages of CAP patients who die within 30 days and one year of discharge without hospital readmission. 4. to estimate the percentage of unplanned readmissions for patients hospitalized with CAP using a structured retrospective chart review. 5. to describe pneumonia-related and -unrelated reasons for readmission for patients hospitalized with CAP using a structured retrospective chart review. 6. to quantify the probability that an unplanned readmission is avoidable using latent class analysis based on the independent classifications by four reviewers. 7. to investigate the distribution of potentially avoidable readmission and the time since discharge from index hospital admission. 8. to identify the characteristics derived from administrative claims data that are independently associated with potentially avoidable readmission. 9. to derive and internally validate an administrative claims data-based risk prediction model for identifying potentially avoidable 30-day readmission of patients hospitalized with CAP. 10. to identify the characteristics abstracted from medical record that are independently associated with potentially avoidable readmission. 11. to derive and internally validate a medical record data-based risk prediction model for identifying potentially avoidable 30-day readmission of patients hospitalized with CAP. 12. to compare the overall accuracy, discrimination, and calibration for the administrative claims data-based versus medical record data-based risk prediction model for identifying potentially avoidable 30-day readmission of patients hospitalized with CAP. 13. to externally validate published risk prediction models for potentially avoidable 30-day readmission. II. Study design R3P is a retrospective observational cohort study of consecutive adult patients discharged from two hospitals with a diagnosis code of CAP. Data sources include routinely collected hospital discharge data and retrospective chart reviews. III. Participating study centers and setting The study will be conducted in a university-affiliated and a general hospitals in Northern Alps, France. With a capacity of 1,362 acute care beds, Grenoble university hospital serves a predominantly urban population of 675,000 inhabitants and reported 135,999 stays in 2014. Annecy general hospital has a capacity of 896 acute care beds and reported 70,651 stays in 2014. IV. Data collection Two clinical research assistants will perform structured retrospective chart review using a computerized data collection instrument. The following variables will be recorded for index hospitalizations: * patient and hospital stay identifiers; * baseline patient characteristics, including demographics, preexisting comorbid condition, pneumonia severity index (PSI) risk class, physical examination and laboratory findings on admission, X-ray or CT-scan findings within 48 hours of admission, initial microbiological work-up; * inhospital antibiotic therapy and associated treatments; * index hospital admission course (ICU admission, pneumonia-related and -unrelated complications); * physical examination and laboratory findings at discharge; * discharge plan and treatments; The following variables will be recorded for the first readmission within one year of discharge: * patient and hospital stay identifiers; * physical examination and laboratory findings on readmission, X-ray or CT-scan findings within 48 hours of readmission; * ICU admission; * pneumonia-related and -unrelated complications; * primary and secondary reasons for readmissions. As part of the French diagnosis-related group (DRG)-based prospective payment system, computerized hospital discharge data include patient and hospital stay identifiers, admission and discharge dates, age, gender, length of stay, discharge location, primary and secondary ICD-10 discharge diagnosis codes (up to 99) for both index hospitalization and readmission. ICD-10 coding complies with national guidelines and is done by trained technicians or physicians, depending on the hospital. Coders usually abstract diagnoses from physician notes, although admission notes, daily progress notes, consultation reports, diagnostic imaging, and treatments are routinely recorded in the medical chart. Discharge data are externally audited by reabstracting a random sample of hospital stays every year. V. Data management To ensure optimal quality, all data collected retrospectively by chart review will be entered electronically by clinical research assistants using a personal identification code and password-protected web-based data collection system. The clinical research assistants will receive formal training in the methods of data abstraction and recording. An operation manual that includes definitions and acceptable data sources for all variables will be distributed. Reliability of data abstraction will be assessed by randomly selecting cases for independent collection by a practicing physician. VI. Statistical analysis 1. Overview The administrative hospital discharge datasets will be linked to medical record-abstracted cases using patient and hospital stay identifiers. If a patient is admitted more than once during the study period, only the first hospitalization will be included as the index admission. 2. Accuracy of ICD-10 discharge diagnosis codes for CAP The accuracy of ICD-10 discharge diagnosis codes will be assessed using three reference methods: 1. Medical record and/or discharge letter notation of CAP diagnosis. 2. Medical record notation of ≥ 1 respiratory symptom (cough, sputum production, dyspnea, tachypnea, or pleuritic pain), and ≥ 1 auscultation finding (rales or crepitations), and ≥ 1 sign of infection (temperature \> 38°C, shivering, or white blood cell count \>10,000/µL or \<4,000/µL), and a new infiltrate on chest radiography or CT-scan performed within 48 hours of admission. 3. A composite of #a and/or #b Positive predictive value point estimates along with 95% confidence intervals (CI) will be reported for the three reference methods, separately. 3. Assignment of reasons for readmission Four physicians, who are members of a clinical review panel, will independently review medical records for both index hospitalization and readmission. The members of the panel are general internists, infectious or respiratory disease specialists with clinical experience in managing CAP patients. Consistent with Jasti et al. (CID 2008;46:550-6), each physician reviewer will use predefined criteria to categorize the reason for rehospitalization as: 1. pneumonia-related worsening of signs or symptoms; 2. new or worsening comorbid condition(s) independent of pneumonia; 3. any combination of pneumonia-related and comorbidity-related reasons. The reviewers will independently assign the primary reason for readmission, using the mutually exclusive categories published by Halfon et al. (J Clin Epidemiol 2002;55:573-87). Inter-rater agreement for the reason for rehospitalization will be quantified using Cohen's Kappa. 4. Identifying potentially avoidable readmissions Consistent with van Walraven et al. (CMAJ 2011;183:e1067-72), four physician reviewers will use a 6-point ordinal scale to rate whether the readmission is an adverse event and whether the adverse event could be avoided. A readmission with a rating above three in both domains will be classified as potentially avoidable by this reviewer. The probability that a readmission is avoidable will be quantified using latent class analysis based on the independent classification by four reviewers. 5. Development of a medical record data-based risk prediction model We will derive a multivariable logistic regression model for quantifying the likelihood that 30-day readmission is avoidable. The development sample will include all cause 30 days readmissions for adult patients hospitalized with CAP. The outcome of interest is a potentially avoidable readmission within 30 days of discharge. We will use medical record-abstracted data for both the index hospitalization and readmission. Candidate predictors will be identified by combining the findings from a systematic literature review with the opinions of field experts. 6. Development of an administrative claims data-based risk prediction model In a separate analysis, we will derive a multivariable logistic regression model for quantifying the likelihood that 30-day readmission is avoidable. The development sample will include all cause 30 days readmissions for adult patients hospitalized with CAP. The outcome of interest is a potentially avoidable readmission within 30 days of discharge. We will use only administrative data and the model will be based on combinations of diagnostic codes between the index hospitalization and the readmission. Candidate predictors will be identified by reviewing published models in combination with the opinions of field experts. 7. Internal validation of risk-prediction models We will use bootstrapping for quantifying the optimism in the apparent performance of clinical prediction models. Discrimination will be quantified by the concordance (c) statistic. Calibration will be investigated graphically and formally assessed by the calibration slope test. VII. Harms and safety In this retrospective observational study, severe adverse events will be collected through a structured chart review, including inhospital mortality, complications due to CAP or CAP management, new or worsening comorbidity, ICU admission, and out-of-hospital mortality. VIII. Auditing Monitoring of study processes and documents will be conducted by personnel designated by the Direction de la Recherche Clinique et de l'Innovation (DRCI) at Grenoble university hospital.