Unplanned Returns to Hospital Care: A Linked Data Study Kathy SMITH 1 - - PDF document

Unplanned Returns to Hospital Care: A Linked Data Study

Kathy SMITH1 and Renee IANNOTTI Health System Information and Performance Reporting Branch, NSW Ministry of Health

• Abstract. The linkage of data across facilities and settings of care provides a holistic

view of the patient journey through the healthcare system. This study, through data linkage, reviews alternative approaches to the measurement of unplanned returns to care in NSW public hospital emergency departments and admitted patient care

• settings. The study shows that existing measures of unplanned returns do not identify

the true extent of these events and highlight the need to develop new approaches to measurement using the increasing availability of integrated patient information.

• Keywords. Representations, readmissions, data linkage, admitted patients,

emergency departments, journeys of care, measurement, indicators

Introduction The objective of the study was to investigate how the linkage of currently disparate but routinely collected patient data could better inform the understanding of patient’s unplanned returns to care and to demonstrate the potential of using existing data in new ways. For system performance managers the availability of integrated data can overcome known weaknesses of current measures and indicators that rely on restricted views of the patient journey, only taking in activities occurring in a single setting, facility or health

• service. This often leaves measures open to misinterpretation due to those missing data

that may otherwise enrich the view of a patient’s overall healthcare journey. Facilitating linkage between data sources opens up opportunities for new and more meaningful

• measurement. For measures such as admitted patient (AP) readmissions and emergency

department (ED) representations linkage provides the opportunity to explore more realistic views of how patients travel through the healthcare system and the dynamics of how, why, when and where they may return to care. Unplanned representations are measured for a variety of reasons1,2 including patient safety, demand management and general understanding of the dynamics of care. Patients will make unplanned returns to care for many reasons. Of particular interest is when unplanned representations are unexpected, avoidable and unnecessary, however this is often only clear on individual record review. The primary measurement issue for this study was identifying all unplanned returns to care to provide a consistent base from which more targeted investigation could be

• undertaken. Traditionally, available data and systems have only allowed us to separately

1 Corresponding Author.

view these returns to care from within single settings such as EDs or AP settings, and within individual health care services. Significant improvements in the availability and timeliness of linked data have allowed a new view of representations to care that follows patient movements within and across the health services. This study makes a comparison of measures of patients returning to care using linked and unlinked datasets.

• 1. Background

There is substantial difference between the methods used to define, measure and report unplanned readmissions3,4,5,6. Often this is simply because the focus of interest may vary, resulting in different inclusions and exclusion to the base calculations. Definitions may also be adjusted to address limitations of the available data. The result is, that despite being called “readmissions” the comparability of these measures is not straightforward and often they are not comparable at all. Most studies that examine patients making unplanned returns to care have focused

• n a single setting, either EDs7 or AP settings8,9,10, and generally target only those patients

and events that occur within tightly defined timeframes (e.g. 48 hours and 28 days). Often the focus is only a small cohort of patients with targeted conditions both in the index and readmission event. These scope limiters are often used to increase the likelihood of capturing returns to care that are more likely to relate, or have a causal link to the patient’s previous healthcare event. While this may be fit for a particular focused investigation it also removes from visibility many readmissions that should be investigated. In 2014 a revised 28 day all cause unplanned readmission performance indicator was introduced in the NSW public health setting that did not discriminate or attempt to presume the cause of readmission, but simply aimed to identify that an unplanned readmission occurred. This measure however only covers the admitted patient setting and omits the common occurrence of a patient presenting at an ED without having an AP

• event. The recent improvements in the availability of data linkage facilities across NSW

health allowed for a more sophisticated view of patients who readmit as an inpatient and/or represent at an ED following a health care service to be investigated. In NSW 66%11 of public hospital ED presentations during 2015 involved patients departing from the ED without being admitted. Despite this, few investigations have been undertaken on identifying or measuring ED attendances following an AP stay or on AP stays following ED attendances, particularly in an Australian context. A study in the US by Brennen et al12 found 18.2% of patients had an ED visit within 30 days of an AP stay. Dinh et al13 examined readmissions to an AP unit within 30 days of index AP admission from ED as well as unplanned representations to ED within 3 days of discharge from ED. A study by Robinson et al14, in a NSW tertiary level ED found 23.7% of patients who represented to the ED required hospital admission. Related investigations were undertaken and published by the Bureau of Health Information9,10. Many of the previous studies have been limited by being restricted to readmissions to the same facility as the initial presentation. Davies et al15. found that in the US 68% of all-cause readmissions and 70% of 30-day potentially preventable readmissions occurred to the same hospital indicating that at least 30% of readmissions were to a different hospital as the index admission.

• 2. Methods

A retrospective data linkage study was undertaken of all patients presenting to NSW public EDs or discharging from an AP stay at a NSW public hospital between 1 January 2011 and 31 December 2015. The primary data sources were the NSW Admitted Patient Data Collection (APDC) and the NSW Emergency Department Data Collection (EDDC). ED and AP records were probabilistically linked by the Centre for Health Record Linkage16 (CHeReL) using blocking and scoring to identify matches. The reported quality of the record linkage is less than 5 per 1,000 missed and 5 per 1,000 false positive

• links. Patients with an AP stay or ED presentation between 1 January 2011 and 31

December 2015 were followed for 28 days following their discharge from hospital or departure from ED to determine if the patient was readmitted or represented for an unplanned event. Data was extracted from the “Admitted Patient, Emergency Department Attendance and Deaths Register” (APEDDR) via SAPHaRI (Secure Analytics for Population Health Research and Intelligence) which is managed by the Centre for Epidemiology and Evidence, NSW Ministry of Health. There are no indicators in the NSW public health system which report on representations to hospital care by combining AP readmissions and ED representations into a single metric. To enable such analysis the concept of a journey of care (referred to as a journey) is introduced This consists of contiguous hospital events beginning when a patient first interacts with a hospital (either in the ED or AP setting) until the patient completes all hospital events in the contiguous series (i.e. until the patient leaves the care

• f the health system). Some journeys of care may involve multiple contiguous events

within one hospital or across many hospitals such as when a patient is transferred between hospitals. Patients with an event that resulted in death (either ED or AP) were included in counts of representation but were excluded as index events. Descriptive analyses using seven methods of measuring readmissions and representations were performed:

• two based on AP data only (same facility and any facility readmissions);
• two based on ED data only (same facility and any facility ED representations)

and;

• three based on linked AP and ED data.

The two methods using only ED data included ED attendances where the patient was either admitted and discharged as an inpatient in ED or departed treatment completed. Cochrane Armitage trend tests were used to examine the trend in representation rates for each method over the 5 year period between January 2011 and December 2015. A more detailed analysis for a 1 year subset (2015) was undertaken to identify specific

• issues. Analyses were conducted in SAS Enterprise Guide version 6.117.
• 3. Results

There were 17,127,716 journeys of care between 1 January 2011 and 31 December 2015 for 5,057,028 individual patients. 54.5% of journeys involved ED only, 28.6% involved AP only and 16.8% involved both the ED and AP. 2.0% of journeys involved multiple

• facilities. The journeys of care resulted in 2,743,969 unplanned representation journeys

to either an ED or AP setting within 28 days.

Figure 1 displays the monthly readmission/representation rates between January 2011 and December 2015 for each of the various calculation methods. Irrespective of the measure there is a slight but significant increase in unplanned readmission/representation rates over the five years (p-values <0.0001 for all methods except the any facility ED representation (based on ED attendances) method where there was a slight but not significant increase in the unplanned representation rate.

Figure 1. Monthly 28 day unplanned readmission and representation rates using different definitions, NSW public facilities, 2011-2015.

Table 1 presents the annual 28 day unplanned readmission and representation rates using the seven different methods for 2015. The lower rate for the any facility linked journey vs. any facility linked AP only rate is most likely due to the process of linkage changing the denominator values and reduced counting of transfers as discrete admissions, particularly when patients are transferred via the ED to an AP bed. Similarly for ED rates the any facility linked journey data is lower than the any facility linked ED

• nly data for similar reasons. Both linked results however remain higher than rates

derived from the current unlinked ED representation method.

Table 1. Annual 28 day unplanned readmission and representation rates using different definitions, NSW public facilities, 2015. Readmission/representation Method Numerator Denominator Rate Same Facility Readmission (based on inpatient stays) 107,592 1,808,626 5.9% Any Facility Readmission (based on inpatient stays) 132,039 1,808,626 7.3% Any Facility Representation Involving AP (based on journeys) 217,330 3,650,898 6.0% Same Facility ED Representations (based on ED attendances) 238,458 1,689,077 14.1% Any Facility ED Representations (based on ED attendances) 295,656 1,689,077 17.5% Any Facility Representation Involving ED (based on journeys) 562,399 3,650,898 15.4% Any Facility Representation (based on journeys) 589,966 3,650,898 16.2%

The all-cause, any facility representation rate for 2015 was 16.2% (Table 1), higher than both the journeys representation rate based on returns to ED or AP alone. Figure 2 shows representations, for journeys (linked across setting and facility) based on the type

• f index journey of care. These rates show different results again from those that look at

single setting focused data, even when the data is linked. The results again reinforce the importance of measuring unplanned returns to ED as well as AP. Where people represent has been the basis of much speculation over the years. Without linking the data across facilities only those patients that return to the site of

• riginal treatment can be seen. This study shows that the majority of unplanned returns

to care will be to the same facility (71.6%) (Figure 3). However the higher rate of unplanned representations for the linked and journey data compared to unplanned readmissions using AP data only indicates that more patients will first represent to an ED either alone or prior to readmission as an AP (Figure 3). Overall almost a third (28.4%)

• f unplanned representations will be to a different facility than the index facility. And,
• f these up to 18.0% were to a facility in the same health district (Figure 3). There is

however clear variation based on the patient’s area of residence with up to 16.3% of patients in metropolitan areas making unplanned representations to a facility outside the health district altogether while only 4.4% rural patients made unplanned representations

• utside their local health district.

Figure 2. Monthly 28 day unplanned readmission and representation rates using different definitions, NSW, 2011-2015. Figure 3. 28 day unplanned representations to the same/different facilities/LHDs by remoteness

• f

residence and type

• f

representation, 2015.

• 4. Discussion

The results show that very different views of unplanned returns to care can be achieved by using different calculation methods. In particular they demonstrate that measurement based on single settings and/or facilities significantly underestimate the quantum of unplanned returns to care locally within a health service as well as across the broader public health system. The study clearly demonstrates that by examining only unplanned readmissions back to the original admitting setting and ignoring patients representing to the emergency department misses the large proportion of representations to ED which do not result in

• admission. Counting only representations to the same facility, misses the sizeable

percentage of representations which are to a different facility and/or health service. The study confirms and quantifies the long held assumption that patients will make unplanned representations for care to other settings within the same facility and to other facilities within and outside individual local health districts. These results provide a view not currently visible to decision making and evaluation processes. The use of linked and patient journey data presented in this study provides a more holistic view of patient care and encourages the development of system focused solutions and management of issues such as unplanned representation to care.

• 5. Conclusions

This study demonstrates the power of integrated healthcare data used for routine measurement and evaluation to provide new insights into complex healthcare

• environments. As routine health system reporting transitions from reliance on

unconnected administrative data sources to integrated patient based collection and storage systems, the introduction of new approaches to measurement and monitoring metrics harnessing these data from across the system are now or soon will be possible. Current local, state and national routine reporting and analysis is largely restricted to identifying representations made to the same facility and looking at the ED and AP settings independently. With 28% of unplanned representations being to a different facility than the index discharging facility, there is currently a significant portion of unrecognized/unaccounted activity that is excluded from decision making. The current data environment is now maturing and becoming more interconnected with increasing availability of data that can be linked across healthcare events either in the source or local and state data warehouses and repositories. The use of Journeys of Care is an example of leveraging off this environment to better understand our health services. Readmissions are a mainstay indicator for most health service monitoring and

• evaluations. As this study reveals without reference to activity in the alternate

destinations patients may use when making an unplanned returns to care there is likely to be a significant underestimation and/or miscalculation of the true impact of unplanned returns to care on the system as a whole. Discrepancies as are highlighted between the current unlinked measures of unplanned returns to care and those data based on linked and journey based metrics raise immediate questions regarding the importance and meaning of current indicators and how and when to make a transition to new system focused measurement. For State and National reporting there remain a number of technical barriers to be overcome prior to routine reporting of linked patient journeys. At State level using linkage facilities like the CHeReL, whole of state perspectives can be

investigated for future reporting, policy and service model development. At the local health service level integration of the patients care records at source should allow for immediate review of patient returns to care across the AP and EDs at least at facility level and across the whole health service where a single patient record and/or identifier is used. The results reinforce that actions in one area of the health system are seldom independent or without impact on other areas and that the data on which we base decisions should reflect this. References

[1] Kristensen S R, Bech M, Quentin W, A roadmap for comparing readmission policies with application to Denmark, England, Germany and the United States, Health Policy 119 (2015) 264-273 [2] White S, Moola S, McMillan M, Riitano D, Aromataris E, Unplanned hospital readmission rates and their use in clinical practice and health service management: a rapid review, The Joanna Briggs Institute and The University of Adelaide on behalf of NSW Ministry of Health (unpublished research project) [3] Mulligan T, Chen C, Working Paper: Simple Methods of Measuring Hospital Readmission Rates, March 2012, AHIP Centre for Policy and Research [4] AHRQ Centers for Medicare and Medicaid Services, National Hospital Inpatient Quality Measures, Unplanned readmission: Hospital-wide all-cause, unplanned readmission rate (HWR), Oct 2015 [5] Bradley E, Yakusheva O, Horwitz L I, Sipsma H, Fletcher J, Identifying Patients at Increased Risk for Unplanned Readmission, Med Care. 2013 Sep; 51(9): 761-766 [6] Yam C HK, Wong E LY, Chan F WK, Wong F YY, Leung M CM, Yeoh EK, Measuring and preventing potentially avoidable hospital readmissions: a review of the literature, Hong Kong Med J 2010;16:383-9 [7] Bottle A, Aylin P, Majeed A, Identifying patients at high risk of emergency hospital admissions: a logistic regression analysis, J R Soc Med 2006; 99:406-414 [8] McLean R, Mendis K, Canalese J, A ten-year retrospective study of unplanned readmission to a regional Australian hospital, Aust Health Rev 2008: 32(3): 537-547 [9] Bureau of Health Information. The Insights Series: Return to acute care following hospitalisation, Insights into readmissions, NSW public hospitals, July 2009 – June 2012 Sydney (NSW); BHI; 2015. [10] Bureau of Health Information. The Insights Series – Return to acute care following hospitalisation, eight clinical conditions, NSW, July 2012 – June 2015. Sydney (NSW); BHI; 2017. [11] Bureau of Health Information, Healthcare Observer, http://www.bhi.nsw.gov.au/Healthcare_Observer/ _nocache (access 1 February 2017). [12] Brennan JJ, Chan TC, Killeen JP, Castillo EM, Inpatient Readmissions and Emergency Department Visits within 30 Days of a Hospital Admission, West J Emerg Med. 2015;16(7):1025-1029. [13] Dinh MM, Berendsen Russell S, Bein KJ, Chalkley D, Muscatello D, Paoloni R, Ivers R, Trends and characteristics of short-term and frequent representations to emergency departments: A population-based study from New South Wales, Australia, Emergency Medicine Australasia, 2016, 28: 307-312. [14] Robinson KK and Lam B, Early emergency department representations, Emergency Medicine Australasia, 2013, 25: 140-146.S. [15] Davies, O. Saynina, K. McDonald and L. Baker, “Limitations of using same-hospital readmission metrics,” International Journal for Quality in Health Care, 2013. [16] Centre for Health Record Linkage: CHeReL Quality Assurance. http://www.cherel.org.au/quality- assurance (accessed 3 February 2017). [17] SAS Institute, Cary, NC, USA.