Article Text
PDF
Abstract
Background Avoidable hospital readmission is a focus of quality improvement efforts. The effectiveness of individual elements of the standard discharge process in reducing rehospitalisation is unknown.
Methods The authors conducted a case-control study of 1039 patients experiencing rehospitalisation within 30 days of discharge and 981 non-rehospitalised patients matched on admission diagnosis, discharge disposition, and severity of illness. In separate models for each discharge process component, the authors measured the relationship between readmission and discharge summary completion, contents of discharge summary, completion of discharge instructions, contents of discharge instructions, presence of caregiver for discharge instruction, completion of medication reconciliation, and arrangement of ambulatory follow-up prior to discharge.
Results Adjusting for patient and hospital characteristics, including severity of illness and discharge disposition, the study failed to find an association between readmission and most components of the discharge process. There was no association between readmission and medication reconciliation, transmission of discharge summary to an outpatient physician, or documentation of any specific aspect of discharge instruction. Associations were found between readmission and discharge with followup arranged (adjusted odds ratio (OR) 1.21; 95% CI 1.05 to 1.37) and increasing number of medicines (adjusted OR 1.02; 95% CI 1.01 to 1.04).
Conclusions Documentation of discharge process components in the medical record may not reflect actual discharge process activities. Alternatively, mandated discharge processes are ineffective in preventing readmission. The observed absence of an association between discharge documentation and readmission indicates that discharge quality improvement initiatives should target metrics of discharge process quality beyond improving rates of documentation.
Statistics from Altmetric.com
Hospital readmission occurs within 30 days of hospital discharge for one in five (19.6%) older adults in the US Medicare public insurance program, accounting for an excess healthcare cost of US$12 billion annually.1 2 As a result, American healthcare reform legislation has targeted avoidable rehospitalisation as a focus for quality improvement and authorises reimbursement adjustments to hospitals based on rehospitalisation rates. As these incentives are implemented, hospitals will likely face growing pressure to enhance processes to reduce the likelihood of readmission.
Regulatory mandates for discharge process documentation for American hospitals include a discharge summary documenting diagnosis, findings and interventions; discharge instructions for the patient or caregiver; and medication reconciliation that requires a patient's discharge medications to be compared with previous home medications and differences reconciled. Despite mandates, these discharge process elements are frequently not conducted and up to one-third of patients feel their post-discharge care plan is inadequate.3–11 Furthermore, Jha et al failed to find a significant relationship between documented delivery of discharge instructions and readmission in a sample of a limited clinical and demographic population.7 Little is known about the relationship between additional mandatory components of discharge documentation such as medication reconciliation and the contents of the discharge summary and likelihood of readmission. In order to increase understanding of the role of discharge process elements in reducing hospital readmission, the authors evaluated the relationship between elements of the discharge process and readmission among 34 hospitals within a national not-for-profit, membership-based hospital benchmarking organisation, the University HealthSystem Consortium (UHC).
Methods
Study design
In May 2009, 34 participating academic health centres collected data for the UHC Reducing Readmissions benchmarking project. UHC is composed of 107 academic medical centres and 232 of their affiliated hospitals. UHC benchmarking efforts include longitudinal medical chart abstraction on topics related to the quality of healthcare.
Patients were eligible for inclusion in the study if the principal diagnosis associated with the index admission was one of nine diagnostic groups at high risk for readmission (box 1).12 Diagnostic groups 1–3 (acute myocardial infarction, heart failure, pneumonia) reflected those diagnoses used in the Centers for Medicare and Medicaid Services risk-adjusted readmission public reporting initiative and were limited to adult inpatients ≥65 years of age who were insured with Medicare; otherwise adult inpatients ≥18 years of age with any payer status were eligible for inclusion. Exclusion criteria were expired during index admission, transferred to another acute care, and discharged against medical advice. Patients with myocardial infarction for whom readmission within 30 days of discharge was for planned revascularisation procedures or who were discharged alive on the day of readmission were also excluded. Index admission or readmission episodes with a principal procedure code for chemotherapy, radiation, rehabilitation, dialysis, or obstetrical delivery were also excluded. The number of staffed beds and the geographical region of the hospital were recorded. Inclusion and exclusion criteria were selected in order to define a cohort of adults at moderate to increased risk of 30-day rehospitalisation.
Diagnostic inclusion criteria for patient groups
AMI (using CMS specifications for 30-day, risk-standardised readmission measures 2008 dry run (ICD-9-CM code 410.xx)).
HF (using CMS specifications for 30-day, risk-standardised readmission measures 2008 dry run (ICD-9-CM codes 402.01, 402.11, 402.91, 404.01, 404.91, or 428)).
PN (using CMS specifications for 30-day, risk-standardised readmission measures 2008 dry run (ICD-9-CM codes 480.0–480.3, 480.8, 480.9, 481, 482.0–482.3, 482.31, 482.32, 482.39, 482.40, 482.41, 482.49, 482.81, 482.81–84, 482.89, 482.9, 483.0, 483.1, 483.8, 485, 486, 487.0)).
COPD (APR-DRG 140).
Psychological (APR-DRG 750, 751, 753).
Renal failure (APR-DRG 460).
Septicemia (APR-DRG 720).
Bowel procedure (APR-DRG 221).
Orthopaedic (APR-DRG 301, 302, 308).
UHC staff identified cases and controls for the study from the UHC Clinical Database (CDB), a data repository maintained by UHC and composed of patient-level clinical data from member organisations. Cases were enrolled in reverse chronological order by discharge date, beginning with eligible discharges on 31 December 2008 and proceeding back in time without skipping any eligible case based on inclusion and exclusion criteria, until 30 cases experiencing same-hospital readmission were enrolled from each participating hospital. Matching of cases and controls was conducted at the hospital level based on admission diagnosis, discharge disposition, and severity of illness. Severity of illness was assigned based on All Patient Refined Diagnosis Related Groups.13
Patient data were taken from the UHC CDB or were collected by chart abstraction by staff with previous UHC-related training and experience in medical record data abstraction. Data were entered into an electronic data entry tool. Abstractors were not blinded to the readmission outcome. Project staff reviewed index admission records for patient demographics (age, gender and race/ethnicity) and for the following components of the discharge process: discharge summary (presence of summary within 7 days of discharge and documentation of reason for hospitalisation, hospital course, procedures, results of procedures, plan for follow-up and full medication list), documentation that discharge summary was forwarded to an outpatient physician, documentation of delivery of discharge instructions to the patient and/or caregiver (including the following: full medication list, instructions on when to hold medications, symptoms requiring medical attention, discharging physician contact information), medication reconciliation and number of discharge medications, presence of scheduled medical follow-up at the time of discharge.
Chart abstractors received formal live training and were given rigorous definitions for each data element as part of a standard variable guide shared with all sites. During data collection a clarification document was maintained to provide a reference based on areas of uncertainty raised by chart abstractors. Every data collection form was examined for data entry errors and omissions. When illogical entries or missing data were encountered, the data collection forms were returned to the member institutions for clarification or correction.
Data analysis
This study modelled rehospitalisation (yes vs no) as the outcome of interest. Independent variables included presence and elements of discharge instructions, presence and elements of the discharge summary, presence of scheduled follow-up appointment at discharge, number of medicines on discharge, and documentation of completed medication reconciliation. Bivariate testing for differences between the readmission cohort and the non-readmission cohort was conducted with χ2 for categorical predictors and Student t test for continuous predictors. In multivariate models, generalized estimating equation models with an exchangeable correlation structure were used to account for hospital level effects on the relationship between discharge processes and readmission.14 15 A separate generalized estimating equation model was fit for each discharge process element. Both unadjusted models and adjusted models were estimated accounting for patient race/ethnicity, age, gender, severity of illness, diagnostic category, discharge disposition and specific hospital characteristics, including admitting hospital, number of staffed beds and admitting hospital region. Unadjusted and adjusted odds ratios (ORs) and 95% CIs were calculated. The number of medications prescribed at discharge in the readmission cohort were compared with the non-readmission cohort using Student t test. Data were analysed using SAS version 9.2.
Results
Within the CDB, 1020 readmission cases and 1014 matching controls were identified. During full chart review, 29 records were reallocated based on a readmission outcome misidentified in the CDB and 14 records were excluded based on exclusion criteria. The final readmission cohort consisted of 1039 readmission cases and 981 controls. There were no significant differences in the designated patient-level matching variables between the cohort of patients readmitted to hospital and the control group not readmitted (table 1). The study did identify significant differences in payer status between the cohorts, with patients who were readmitted being more likely to be insured and to have public insurance (US Medicare or Medicaid programs) instead of private (employer-based or individually purchased) insurance.
Hospital (n=34) and patient (n=2020) sample characteristics
Characteristics of discharge instructions
Absence of discharge instructions in the medical record was not more common among patients in the readmission cohort compared with those in the non-readmission cohort (table 2). Odds of readmission were also not significantly associated with documentation of a full medication list, instructions on when to hold medications, symptoms requiring medical attention, or contact information for the discharging physician. Documentation of delivery of discharge instructions to a family member showed an association with not being rehospitalised, which approached statistical significance at the 0.05 level (adjusted OR 0.60; 95% CI 0.36 to 1.01).
Frequency of discharge process elements, unadjusted and adjusted* associations between discharge process elements and readmission cohort (readmission cohort compared with non-readmission cohort)
Characteristics of discharge summary
Patients whose records included a discharge summary dated within 7 days of discharge were equally likely to be readmitted as patients without a timely discharge summary. There was also no association between documentation of transmission of the discharge summary to an outpatient provider and readmission. No mandatory elements of the discharge summary (including reason for hospitalisation, hospital course, procedures and results of procedures, plan for follow-up care, full medication list) were associated with a lower likelihood of readmission.
Number of discharge medications and medication reconciliation
A modest association was identified between the number of medications on discharge and the likelihood of subsequent readmission in adjusted models (adjusted OR 1.02; 95% CI 1.01 to 1.04). This reflected a 2% increase in readmission risk for each additional medication present upon discharge while accounting for severity of illness. In an unadjusted analysis, readmitted patients were found to report significantly fewer medicines in the readmission medication list than were present on the discharge medication list (mean 10.85 (SD 6.19) compared with 10.40 (6.24)). The relationship was also assessed between change in the number of medications between index admission and discharge and the likelihood of subsequent readmission, but no significant association was found. In addition, no significant association was found between documentation of medication reconciliation and subsequent readmission.
Presence of scheduled medical follow-up
The study identified a positive and significant association between documentation of a scheduled follow-up appointment at the time of discharge and subsequent readmission (adjusted OR 1.21; 95% CI 1.05 to 1.37).
Discussion
The study results fail to demonstrate significant associations between documentation of mandatory discharge elements and likelihood of 30-day hospital readmission. Possible explanations for this finding, including inaccurate documentation and ineffective discharge processes, are concerning. The possibility that these discharge elements are not protective is consistent with concerns raised by Jha and others about the effectiveness of discharge process elements.7 11 Currently these activities represent the principal focus of provider effort dedicated to the transition from hospital to the post-acute care setting and thus should have proven effectiveness.
Because our data represent documentation of processes rather than direct observation of the conduct of these activities, our results may reflect poor quality of discharge processes which were nonetheless documented as complete. Documentation in the medical record may simply reflect an attempt to comply with regulatory standards—a ‘documented versus done’ differential—and not an accurate measure of the quality of the effort. Similar to previous studies,5 the authors identified incomplete documentation of mandatory elements of the discharge process. This incomplete documentation may represent a lack of engagement in the discharge process, which may be manifest in the quality of discharge processes documentation.
The study findings should be considered in light of several limitations. The effect of some discharge process elements may be masked by indication bias despite adjustment for severity of illness. For example, a significant positive association was identified between establishment of a follow-up appointment and subsequent hospital readmission. However, the multivariate analysis did adjust for severity of illness in an attempt to mitigate this effect. In addition, because patient record review was limited to the index hospital, readmission to other institutions was not recorded. However, these unmeasured readmissions are unlikely to have meaningfully different index discharge process characteristics and contribute to a type II error. Several of the models contain broad CIs, which are thought to be primarily because the models accounted for clustering of patients at the hospital level rather than underpowered analysis. To test logistic regression, models were also constructed that did not account for clustering effects, which reflected more precise intervals.
Whether the absence of an association between documentation of a ‘complete’ discharge process and protection against readmission is the result of failure to document actual events or failure of process design is unclear from this study. Further research is needed to assess the absolute quality of discharge components that are documented as complete but may be inadequately performed. Given the unclear effect of common discharge processes, additional studies that clarify the effectiveness of these components of discharge are necessary.
Recently enacted US healthcare reform places responsibility for reducing readmission squarely on hospitals. This will likely shift the focus of management of care transitions away from documentation and towards efforts to reduce the readmission outcome itself. As hospitals increase efforts to reduce readmission, tasks that do not reduce the likelihood of readmission should be replaced with effective processes.
References
Footnotes
Presented at the Annual Meeting of the Society for Hospital Medicine. Washington, DC, 8 April 2010.
Competing interests None.
Provenance and peer review Not commissioned; externally peer reviewed.