Discharge status of the patient: evaluating hospital data quality with a focus on long-term and palliative care patient data.

Abstract

Dear Editor, Health administrative data, as found in hospital morbidity datasets are valuable data sources that inform epidemiological studies such as the Global Burden of Disease study (GBD 2019 Diseases and Injuries Collaborators, 2020), and can be used to achieve many aims in relation to health services research and management. Furthermore, Diagnosis Related Group (DRG) systems rely on administrative data, namely diagnosis/procedure codes, age, sex, and discharge destination (Averill et al., 2003) and in many countries are used for hospital reimbursement purposes (Geissler et al., 2011; Mathauer and Wittenbecher, 2013). In this context, the quality of health records, which constitutes the basis for the construction of administrative datasets through clinical coding (Alonso et al., 2020), is paramount. Clinical coding quality issues have been widely discussed (Cheng et al., 2009; Dafny, 2005; O’Malley et al., 2005; Pongpirul and Robinson, 2013; Southern et al., 2015), but little attention has been paid to issues associated with some administrative variables, such as discharge destination, despite their potential impact on the financial reimbursements received by hospitals, as previously mentioned in the case of Medicare (Centers for Medicare & Medicaid Services, 2018). Presented in this letter is our analysis of the quality of this variable, which is essential for DRG grouping and can also be reused for many other purposes. Discharge destination, as a variable, is currently categorised according to standard codes, using information abstracted from hospital documentation. We assessed 2016 data from the Portuguese Hospital Morbidity Database (HMD), which includes administrative data collected from all mainland public and public–private partnership hospitals (62 institutions). These data, described as hospital discharges, were compared to referrals to long-term and palliative care as recorded in the National Network for Long-Term Care (Rede Nacional de Cuidados Continuados Integrados – RNCCI) dataset. RNCCI data are obtained from GestCare, an information system that is used to record all RNCCI-related procedures, including referral. RNCCI data, as accessed through the Portuguese National Health Service Transparency Portal (Ministry of Health, 2016), will be therefore mentioned as referrals. As more than 90% of the referrals originate from hospitals (ACSS, 2017), data from the HMD should correspond with this data source. We focused on the quality of data related to hospital referral for long-term and palliative care, which in Portugal is overseen by the RNCCI (D.R., 2006; Lopes et al., 2018). From the variable ‘discharge destination’, discharges to long-term and palliative care categories were chosen due to data availability, as these were the only data categories that have a secondary information source, with which comparisons can be made. We selected HMD categories ‘63 –Discharge to long-term inpatient care’ and ‘51 – Discharge to Palliative Care Unit’, two categories included in RNCCI data. The RNCCI encompasses several types of inpatient and outpatient units, including longterm and maintenance units (Unidades de Longa Duração e Manutenção – ULDM) for individuals with an expected length of stay of 90 or more consecutive days, and Palliative Care Units (Unidades de Cuidados Paliativos – UCP), which are dedicated to individuals with terminal illnesses requiring late stage and end-of-life care. Other RNCCI-related categories were excluded from analysis as they include referrals to different units (e.g., stroke units) and were not available for the variable discharge destination from the HMD. To estimate the correlation, reliability and agreement between the HMD and RNCCI data, we calculated the Spearman’s rank order correlation, the intraclass correlation coefficient and the information-